74
Workshop Organisers John Carroll University of Sussex, UK Nelleke Oostdijk University of Nijmegen, The Netherlands Richard Sutcliffe University of Limerick, Ireland Workshop Programme Committee Rens Bod Universities of Leeds and Amsterdam Ted Briscoe University of Cambridge John Carroll University of Sussex Anette Frank DFKI, Saarbr¨ ucken Gregory Grefenstette Clairvoyance, Pittsburgh Claire Grover University of Edinburgh Sadao Kurohashi University of Tokyo Stephan Oepen CSLI, Stanford University Nelleke Oostdijk University of Nijmegen Richard Sutcliffe University of Limerick Atro Voutilainen Conexor oy i

W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Workshop Organisers

JohnCarroll Universityof Sussex, UKNellekeOostdijk Universityof Nijmegen,TheNetherlandsRichardSutcliffe Universityof Limerick, Ireland

Workshop ProgrammeCommittee

RensBod Universitiesof LeedsandAmsterdamTedBriscoe Universityof CambridgeJohnCarroll Universityof SussexAnetteFrank DFKI, SaarbruckenGregoryGrefenstette Clairvoyance,PittsburghClaireGrover Universityof EdinburghSadaoKurohashi Universityof TokyoStephanOepen CSLI, StanfordUniversityNellekeOostdijk Universityof NijmegenRichardSutcliffe Universityof LimerickAtro Voutilainen Conexor oy

i

Page 2: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

The Workshop Programme

08:45–09:00WelcomeandIntroduction

09:00–10:30SessionI: GrammarDevelopmentI

TheParallel GrammarProjectMiriam Butt, HelgeDyvik, Tracy HollowayKing, Hiroshi Masuichi,ChristianRohrer

TheGrammarMatrix: Anopen-sourcestarter-kit for therapiddevelopmentof cross-linguistically consistentbroad-coverageprecisiongrammarsEmily M. Bender, DanFlickinger, StephanOepen

Parallel distributedgrammarengineeringfor practicalapplicationsStephanOepen,Emily M. Bender, Uli Callmeier, DanFlickinger, MelanieSiegel

10:30–11:00Break

11:00–12:30SessionII: GrammarDevelopmentII

A developmentenvironmentfor large-scalemulti-lingualparsingsystemsHisamiSuzuki

Adaptingexistinggrammars: TheXLEapproachRonaldM. Kaplan,Tracy HollowayKing, JohnT. Maxwell III

Copingwith problemsin grammarsautomaticallyextractedfromtreebanksCarlosA. Prolo

12:30–13:30Lunch

13:30–14:30SessionIII: FormalismsandApproaches

A classificationof grammardevelopmentstrategiesAlexandraKinyon,CarlosA. Prolo

Encodingandreusinglinguistic informationexpressedbyLinguisticPropertiesCarolineHagege,GabrielG. Bes

14:30–15:30PanelSession

Howdoesaformalisminfluencegrammarengineering?HPSG,LFG,LTAG& TheRest

15:30–16:00Break

16:00–17:00SessionIV: Evaluation

Grammarandlexiconin therobustparsingof Italian: Towardsanon-naıveinterplayRobertoBartolini, AlessandroLenci,SimonettaMontemagni,Vito Pirrelli

Machinetranslationasa testbedfor multilingualanalysisRichardCampbell,CarmenLozano,JessiePinkham,MartineSmets

17:00–17:30DiscussionandClosingRemarks

ii

Page 3: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Tableof Contents

TheParallel GrammarProjectMiriam Butt, HelgeDyvik, Tracy HollowayKing, Hiroshi Masuichi,ChristianRohrer . . . . 1

TheGrammarMatrix: An open-sourcestarter-kit for therapiddevelopmentof cross-linguistically consistentbroad-coverageprecisiongrammarsEmily M. Bender, DanFlickinger, StephanOepen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8

Parallel distributedgrammarengineeringfor practicalapplicationsStephanOepen,Emily M. Bender, Uli Callmeier, DanFlickinger, MelanieSiegel . . . . . . . 15

A developmentenvironmentfor large-scalemulti-lingualparsingsystemsHisamiSuzuki . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22

Adaptingexistinggrammars: TheXLEapproachRonaldM. Kaplan,Tracy HollowayKing, JohnT. Maxwell III . . . . . . . . . . . . . . . . . . . . . . . . 29

Copingwith problemsin grammarsautomaticallyextractedfromtreebanksCarlosA. Prolo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

A classificationof grammardevelopmentstrategiesAlexandraKinyon,CarlosA. Prolo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Encodingandreusinglinguistic informationexpressedbyLinguisticPropertiesCarolineHagege,GabrielG. Bes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Grammarandlexiconin therobustparsingof Italian: Towardsa non-naıveinterplayRobertoBartolini, AlessandroLenci,SimonettaMontemagni,Vito Pirrelli . . . . . . . . . . . . . .57

Machinetranslationasa testbedfor multilingualanalysisRichardCampbell,CarmenLozano,JessiePinkham,MartineSmets. . . . . . . . . . . . . . . . . . . 64

iii

Page 4: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Author Index

Bes,GabrielG. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50Bartolini, Roberto . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Bender, Emily M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 15Butt, Miriam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Callmeier, Uli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Campbell,Richard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64Dyvik, Helge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Flickinger, Dan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 15Hagege,Caroline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50HollowayKing, Tracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 29Kaplan,RonaldM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Kinyon. Alexandra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Lenci,Alessandro. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Lozano,Carmen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64Masuichi,Hiroshi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1Maxwell, JohnT. III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Montemagni,Simonetta. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Oepen,Stephan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 15Pinkham,Jessie. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Pirrelli, Vito . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .57Prolo,CarlosA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36,43Rohrer, Christian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Siegel,Melanie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15Smets,Martine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Suzuki,Hisami . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

iv

Page 5: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

The Parallel Grammar Project

Miriam ButtCent.for ComputationalLinguistics

UMISTManchesterM60 [email protected]

HelgeDyvikDept.of LinguisticsUniversityof Bergen

[email protected]

Tracy Holloway KingPaloAlto ResearchCenterPaloAlto, CA 94304USA

[email protected]

Hir oshi MasuichiCorporateResearchCenter

Fuji XeroxCo.,Ltd.Kanagawa259-0157,JAPAN

[email protected]

Christian RohrerIMS UniversitatStuttgart

[email protected]

AbstractWe report on the Parallel Grammar (ParGram)project which usesthe XLE parserand grammardevelopmentplatform for six languages:English,French,German,Japanese,Norwegian,andUrdu.1

1 Intr oductionLarge-scalegrammardevelopmentplatformsareex-pensiveandtime consumingto produce.As such,adesideratumfor theplatformsis a broadutilizationscope.A grammardevelopmentplatformshouldbeableto beusedto write grammarsfor awidevarietyof languagesanda broadrangeof purposes.In thispaper, wereportontheParallelGrammar(ParGram)project(Butt etal.,1999)whichusestheXLE parserandgrammardevelopmentplatform (Maxwell andKaplan,1993) for six languages:English,French,German,Japanese,Norwegian,and Urdu. All ofthe grammarsuse the Lexical-FunctionalGram-mar(LFG) formalismwhichproducesc(onstituent)-structures(trees)andf(unctional)-structures(AVMs)asthesyntacticanalysis.

LFG assumesa versionof Chomsky’s UniversalGrammarhypothesis,namelythatall languagesarestructuredby similar underlyingprinciples.WithinLFG, f-structuresare meantto encodea languageuniversal level of analysis, allowing for cross-linguisticparallelismatthis levelof abstraction.Al-thoughtheconstructionof c-structuresis governed

1We would like to thankEmily Bender, Mary Dalrymple,andRonKaplanfor helpwith thispaper. In addition,wewouldlike to acknowledgethe other grammarwriters in the Par-Gramproject,bothcurrent:StefanieDipper, Jean-PhilippeMar-cotte,TomokoOhkuma,andVictoriaRosen;andpast:CarolineBrun,ChristianFortmann,AnetteFrank,JonasKuhn,VeronicaLux, Yukiko Morimoto, Marıa-EugeniaNino, andFrederiqueSegond.

by generalwellformednessprinciples,this level ofanalysisencodeslanguageparticulardifferencesinlinearwordorder, surfacemorphologicalvs.syntac-tic structures,andconstituency.

TheParGramprojectaimsto testtheLFGformal-ismfor its universalityandcoveragelimitationsandto seehowfar parallelismcanbemaintainedacrosslanguages.Wherepossible,the analysesproducedby the grammarsfor similar constructionsin eachlanguageare parallel. This has the computationaladvantagethat the grammarscanbe usedin simi-lar applicationsandthatmachinetranslation(Frank,1999)canbesimplified.

Theresultsof theprojectto dateareencouraging.Despitedifferencesbetweenthelanguagesinvolvedandtheaimsandbackgroundsof theprojectgroups,theParGramgrammarsachieveahighlevelof paral-lelism.Thisparallelismappliesto thesyntacticanal-ysesproduced,aswell asto grammardevelopmentitself: the sharingof templatesand featuredecla-rations,the utilization of commontechniques,andthetransferof knowledgeandtechnologyfrom onegrammarto another. Theability to bundlegrammarwriting techniques,suchastemplates,into transfer-able technologymeansthat new grammarscanbebootstrappedin a relativelyshortamountof time.

There are a numberof other large-scalegram-marprojectsin existencewhich we mentionbrieflyhere.TheLS-GRAM project(Schmidtetal.,1996),funded by the EU-Commissionunder LRE (Lin-guistic ResearchandEngineering),wasconcernedwith thedevelopmentof grammaticalresourcesfornine Europeanlanguages:Danish,Dutch, English,French,German,Greek, Italian, Portuguese,andSpanish.The project startedin January1994 andendedin July 1996.Developmentof grammaticalresourceswascarriedout in the frameworkof the

1

Page 6: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

AdvancedLanguageEngineeringPlatform(ALEP).Thecoverageof thegrammarsimplementedin LS-GRAM was,however, muchsmallerthanthe cov-erageof the English (Riezleret al., 2002)or Ger-mangrammarin ParGram.An effort whichis closerin spirit to ParGramis the implementionof gram-mardevelopmentplatformsfor HPSG.In theVerb-mobilproject(Wahlster, 2000),HPSGgrammarsforEnglish,German,andJapaneseweredevelopedontwo platforms:LKB (Copestake,2002)andPAGE.ThePAGE system,developedandmaintainedin theLanguageTechnologyLab of theGermanNationalResearchCenter on Artificial Intelligence DFKIGmbH,is anadvancedNLP coreenginethatfacili-tatesthedevelopmentof grammaticalandlexicalre-sources,building on typedfeaturelogics.To evalu-atethe HPSGplatformsandto comparetheir mer-its with thoseof XLE andtheParGramprojects,onewould haveto organizea specialworkshop,partic-ularly as the HPSGgrammarsin Verbmobil werewritten for spokenlanguage,characterizedby shortutterances,whereastheLFG grammarsweredevel-opedfor parsingtechnicalmanualsand/ornewspa-per texts.Therearesomeindicationsthat the Ger-manandEnglishgrammarsin ParGramexceedtheHPSGgrammarsin coverage(see(Crysmannetal.,2002)on theGermanHPSGgrammar).

This paperis organizedasfollows. We first pro-vide a historyof theproject.Then,we discusshowparallelismis maintainedin theproject.Finally, weprovidea summaryanddiscussion.

2 Project History

TheParGramprojectbeganin 1994with threelan-guages:English,French,and German.The gram-marwritersworkedcloselytogetherto solidify thegrammaticalanalysesandconventions.In addition,asXLE wasstill in development,its abilities grewasthesizeof thegrammarsandtheirneedsgrew.

After the initial stageof the project, more lan-guageswere added.BecauseJapaneseis typolog-ically very different from the initial three Euro-peanlanguagesof theproject,it representedachal-lengingcase.Despitethis typologicalchallenge,theJapanesegrammarhasachievedbroadcoverageandhigh performancewithin a year and a half. TheSouthAsian languageUrdu alsoprovidesa widelyspoken,typologicallydistinctlanguage.Althoughitisof Indo-Europeanorigin, it sharesmanycharacter-isticswith Japanesesuchasverb-finality, relativelyfree word order, complexpredicates,and the abil-

ity to drop any argument(rampantpro-drop).Nor-wegianassumesa typologicalmiddle position be-tweenGermanandEnglish,sharingdifferentprop-ertieswith eachof them.BoththeUrduandtheNor-wegiangrammarsarestill relativelysmall.

Each grammarproject has different goals,andeachsite employsgrammarwriters with differentbackgroundsandskills. The English,German,andJapaneseprojects have pursuedthe goal of hav-ing broadcoverage,industrialgrammars.TheNor-wegianand Urdu grammarsare smallerscalebutareexperimentingwith incorporatingdifferentkindsof information into the grammar. The Norwegiangrammarincludesa semanticprojection;theiranal-ysesproducenot only c- andf-structures,but alsosemanticstructures.TheUrdu grammarhasimple-menteda level of argumentstructureand is test-ing various theoreticallinguistic ideas.However,evenwhenthegrammarsareusedfor differentpur-posesand have different additional features,theyhavemaintainedtheir basicparallelismin analysisandhaveprofitedfrom thesharedgrammarwritingtechniquesandtechnology.

Table(1)showsthesizeof thegrammars.Thefirstfigure is thenumberof left-handsidecategoriesinphrase-structureruleswhich compileinto a collec-tion of finite-statemachineswith the listednumberof statesandarcs.

(1)Language Rules States ArcsGerman 444 4883 15870English 310 4935 13268French 132 1116 2674Japanese 50 333 1193Norwegian 46 255 798Urdu 25 106 169

3 Parallelism

Maintainingparallelismin grammarsbeingdevel-opedat differentsiteson typologicallydistinct lan-guagesby grammarwriters from different linguis-tic traditionshasprovensuccessful.At projectmeet-ingsheldtwiceayear, analysesof samplesentencesarecomparedandanydifferencesarediscussed;thegoalis to determinewhetherthedifferencesarejus-tified or whetherthe analysesshould be changedto maintain parallelism.In addition, all of the f-structurefeaturesandtheirvaluesarecompared;thisnot only ensuresthat trivial differencesin namingconventionsdonotarise,butalsogivesanoverviewof theconstructionseachlanguagecoversandhow

2

Page 7: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

theyareanalyzed.All changesareimplementedbe-fore thenextprojectmeeting.Eachmeetingalsoin-volves discussionof constructionswhoseanalysishasnotyetbeensettledon,e.g.,theanalysisof parti-tivesor propernames.If ananalysisis agreedupon,all the grammarsimplementit; if only a tentativeanalysisis found, onegrammarimplementsit andreportsonits success.Forextremelycomplicatedorfundamentalissues,e.g.,how to representpredicatealternations,subcommitteesexaminethe issueandreporton it at the next meeting.The discussionofsuchissuesmaybereopenedatsuccessivemeetingsuntil a concensusis reached.

Evenwithin agivenlinguisticformalism,LFGforParGram,thereis usuallymorethanoneway to an-alyzea construction.Moreover, the sametheoreti-cal analysismayhavedifferentpossibleimplemen-tationsin XLE. Thesesolutionsoftendiffer in effi-ciencyor conceptualsimplicity andoneof thetaskswithin theParGramprojectis to makedesigndeci-sionswhich favor onetheoreticalanalysisandcon-comitantimplementationoveranother.

3.1 Parallel AnalysesWheneverpossible,theParGramgrammarschoosethe sameanalysisand the sametechnicalsolutionfor equivalentconstructions.This was done, forexample,with imperatives.Imperativesarealwaysassigneda null pronominal subjectwithin the f-structureanda featureindicatingthat they areim-peratives,asin (2).

(2) a. Jump! Saute!(French)Spring!(German) Tobe!(Japanese)Hopp! (Norwegian) kuudoo!(Urdu)

b.���� PRED � jump� SUBJ ���SUBJ � PRED � pro��STMT-TYPE imp

��� Another exampleof this type comesfrom the

analysisof specifiers.Specifiersincludemanydif-ferent typesof informationandhencecanbe ana-lyzedin anumberof ways.In theParGramanalysis,thec-structureanalysisis left relativelyfreeaccord-ing to languageparticularneedsandslightly vary-ing theoreticalassumptions.For instance,the Nor-wegiangrammar, unlike the other grammars,im-plementstheprinciplesin (Bresnan,2001)concern-ing therelationshipbetweenanX � -basedc-structureand the f-structure.This allows Norwegianspeci-fiersto beanalyzedasfunctionalheadsof DPsetc.,

whereasthey are constituentsof NPs in the othergrammars.However, at thelevel of f-structure,thisinformation is part of a complex SPEC featureinall the grammars.Thus parallelismis maintainedat the level of f-structureevenacrossdifferentthe-oretical preferences.An exampleis shown in (3)for NorwegianandEnglishin which the SPEC con-sistsof a QUANT(ifier) and a POSS(essive)(SPEC

canalsocontaininformationaboutDETerminersandDEMONstratives).

(3) a. alleminehester(Norwegian)all my horses‘all my horses’

b.�����������PRED � horse�SPEC

������� QUANT � PRED � all ���POSS

��� PRED � pro�PERS 1NUM sg

�� ����� ����������

Interrogativesprovideaninterestingexamplebe-causetheydiffer significantlyin thec-structuresofthe languages,but havethe samebasicf-structure.This contrastcanbe seenbetweentheGermanex-amplein (4) and the Urdu one in (5). In German,the interrogativeword is in first position with thefinite verb second;EnglishandNorwegianpatternlike German.In Urdutheverbis usuallyin final po-sition,but the interrogativecanappearin a numberof positions,includingfollowing theverb(5c).

(4) Was hat JohnMariagegeben? (German)whathasJohnMariagive.PerfP‘What did Johngive to Mary?’

(5) a. jon=nee marii=kookyaadiiyaa? (Urdu)John=Erg Mary=Datwhatgave‘What did Johngive to Mary?

b. jon=neekyaamarii=koodiiyaa?

c. jon=neemarii=kodiiyaakyaa?

Despitethesedifferencesin wordorderandhenceinc-structure,thef-structuresareparallel,with thein-terrogativebeingin a FOCUS-INT andthesentencehavinganinterrogativeSTMT-TYPE, asin (6).

3

Page 8: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(6)����������������

PRED � give� SUBJ,OBJ,OBL ���FOCUS-INT � PRED � pro�

PRON-TYPE int ���SUBJ � PRED � John���OBJ [ ] �OBL � PRED � Mary��STMT-TYPE int

�������������� In the project grammars,many basic construc-

tions are of this type. However, as we will seeinthenextsection,therearetimeswhenparallelismisnot possibleandnot desirable.Evenin thesecases,though, the grammarswhich can be parallel are;so,threeof the languagesmight haveoneanalysis,while threehaveanother.

3.2 Justified DifferencesParallelismisnotmaintainedatthecostof misrepre-sentingthelanguage.Thisis reflectedbythefactthatthec-structuresarenot parallelbecauseword ordervarieswidely from languageto language,althoughtherearenamingconventionsfor thenodes.Instead,thebulkof theparallelismis in thef-structure.How-ever, evenin thef-structure,situationsarisein whichwhatseemsto bethesameconstructionin differentlanguagesdo nothavethesameanalysis.An exam-ple of this is predicateadjectives,asin (7).

(7) a. It is red.

b. Sorewa akai. (Japanese)it TOP red‘It is red.’

In English,the copularverb is consideredthe syn-tacticheadof theclause,with thepronounbeingthesubjectandthepredicateadjectivebeinganXCOMP.However, in Japanese,theadjectiveis themainpred-icate,with thepronounbeingthesubject.As such,thesereceivethenon-parallelanalysesseenin (8a)for Japaneseand(8b) for English.

(8) a.�� PRED � red� SUBJ� �SUBJ � PRED � pro� �

b.��������PRED � be� XCOMP � SUBJ�SUBJ � PRED � pro�� �XCOMP � PRED � red� SUBJ���

SUBJ [ ] � ��������

Another situation that arisesis when a featureor constructionis syntacticallyencodedin onelan-guage,but not another. In suchcases,the informa-tion is only encodedin the languagesthat needit.Theequivalencecapturedbyparallelanalysesisnot,for example,translationalequivalence.Rather, par-allelisminvolvesequivalencewith respectto gram-maticalproperties,e.g.constructiontypes.Onecon-sequenceof this is that a typologically consistentuseof grammaticalterms,embodiedin the featurenames,is enforced.For example,eventhoughthereis a traditionfor referringto thedistinctionbetweenthe pronounshe andsheasa genderdistinction inEnglish,this is a differentdistinctionfrom theonecalled genderin languageslike German,French,Urdu,andNorwegian,wheregenderrefersto nom-inal agreementclasses.Parallelismleadsto thesit-uationwherethe featureGEND occursin German,French,Urdu, and Norwegian,but not in Englishand Japanese.That is, parallelismdoesnot meanfindingthesamefeaturesin all languages,butratherusingthesamefeaturesin thesameway in all lan-guages,to theextentthat theyarejustifiedthere.AFrenchexampleof grammaticalgenderis shownin(9); note that determiner, adjective,and participleagreementis dependenton thegenderof thenoun.Thef-structurefor thenounscrayonandplumeareasin (10)with anovertGEND feature.

(9) a. Le petit crayon estcasse. (French)the-M little-M pencil-M is broken-M.‘The little pencil is broken.’

b. La petite plumeestcassee. (French)the-F little-F pen-F is broken-F.‘The little penis broken.’

(10) ��� PRED � crayon�GEND mascPERS 3

�� ��� PRED � plume�GEND femPERS 3

�� F-structuresfor theequivalentwordsin EnglishandJapanesewill nothavea GEND feature.

A similar examplecomes from Japanesedis-courseparticles.It is well-knownthatJapanesehassyntacticencodingsfor informationsuchashonori-fication. The verb in the Japanesesentence(11a)encodesinformation that the subjectis respected,while the verb in (11b) showspolitenessfrom thewriter (speaker)to the reader(hearer)of the sen-tence.Thef-structuresfor theverbsin (11) areasin

4

Page 9: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(12) with RESPECT andPOLITE featureswithin theADDRESS feature.

(11) a. senseiga honwo oyomininaru.teacherNom bookAcc read-Respect‘The teacherreadthebook.’ (Japanese)

b. seitoga honwo yomimasu.studentNombookAcc read-Polite‘The studentreadsthebook.’ (Japanese)

(12) a.�� PRED � yomu� SUBJ,OBJ���ADDRESS � RESPECT +�

b.�� PRED � yomu� SUBJ,OBJ � �ADDRESS � POLITE +�

A final examplecomesfrom English progres-sives,as in (13). In order to distinguishthesetwoforms,theEnglishgrammarusesa PROG � featurewithin the tense/aspectsystem.(13b) showsthe f-structurefor (13a.ii).

(13) a. Johnhit Bill. i. He cried.ii. He wascrying.

b.��������PRED � cry � SUBJ � �SUBJ � PRED � pro��TNS-ASP � TENSE past

PROG + ��������

However, this distinction is not found in the otherlanguages.For example,(14a) is usedto expressboth(13a.i)and(13a.ii) in German.

(14) a. Er weinte. (German)hecried‘He cried.’

b.����� PRED � weinen� SUBJ ���SUBJ � PRED � pro� �TNS-ASP � TENSE past�

��� As seenin (14b),theGermanf-structureis left un-derspecifiedfor PROG becausethereis no syntacticreflexof it. If sucha featurewereposited,rampantambiguity would be introducedfor all past tenseformsin German.Instead,thesemanticswill deter-minewhethersuchformsareprogressive.

Thus,thereareanumberof situationswherehav-ing parallel analyseswould result in an incorrectanalysisfor oneof thelanguages.

3.3 OneLanguageShowsthe WayAnothertypeof situationariseswhenonelanguageprovidesevidencefor acertainfeaturespaceor typeof analysisthat is neither explicitly mirrored norexplicitly contradictedby anotherlanguage.In the-oretical linguistics, it is commonlyacknowledgedthatwhatonelanguagecodesovertlymaybeharderto detectfor anotherlanguage.This situationhasarisenin theParGramproject.Casefeaturesfall un-der this topic. German,Japanese,and Urdu markNPs with overt casemorphology. In comparison,English,French,andNorwegianmakerelativelylit-tle useof caseexceptaspartof thepronominalsys-tem.Nevertheless,thef-structureanalysesfor all thelanguagescontainacasefeaturein thespecificationof nounphrases.

This“overspecification”of informationexpressesdeeperlinguistic generalizationsand keepsthe f-structuralanalysesasparallelaspossible.In addi-tion, the featurescanbe put to usefor the isolatedphenomenain which theydoplaya role.For exam-ple, Englishdoesnot markanimacygrammaticallyin most situations.However, providing a ANIM +featureto knownanimates,suchaspeople’s namesandpronouns,allowsthegrammarto encodeinfor-mation that is relevantfor interpretation.Considertherelativepronounwho in (15).

(15) a. thegirl[ANIM +] who[ANIM +] left

b. thebox[ANIM +] who[ANIM +] left

TherelativepronounhasaANIM + featurethatisas-signedto thenounit modifiesby therelativeclauserules.As such,a nounmodifiedby a relativeclauseheadedby who is interpretedasanimate.In thecaseof canonicalinanimates,asin (15b),this will resultin a pragmaticallyodd interpretation,which is en-codedin thef-structure.

Teasingapartthesedifferentphenomenacrosslin-guisticallyposesachallengethattheParGrammem-bersarecontinuallyengagedin. As such,we havedevelopedseveralmethodsto help maintainparal-lelism.

3.4 Mechanicsof Maintaining ParallelismTheparallelismamongthegrammarsis maintainedin anumberof ways.Most of thework is donedur-ing two week-longprojectmeetingsheldeachyear.

5

Page 10: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Threemain activitiesoccurduring thesemeetings:comparisonof samplef-structures,comparisonoffeaturesandtheir values,anddiscussionsof neworproblematicconstructions.

A month before each meeting, the host sitechoosesaroundfifteen sentenceswhoseanalysisisto becomparedat themeeting.Thesecanbea ran-domselectionor be thematic,e.g.,all dealingwithpredicativesor with interrogatives.The sentencesarethenparsedby eachgrammarandtheoutputiscompared.For themorerecentgrammars,this maymeanaddingtherelevantrulesto thegrammars,re-sultingin growthof thegrammar;for theoldergram-mars,thismaymeanupdatingaconstructionthathasnotbeenexaminedin manyyears.Anotherapproachthatwastakenat thebeginningof theprojectwastohaveacommoncorpusof about1,000sentencesthatall of thegrammarswereto parse.For theEnglish,French,andGermangrammars,this wasanalignedtractormanual.Thecorpussentenceswereusedforthe initial f-structurecomparisons.Having a com-moncorpusensuredthat thegrammarswould haveroughly the samecoverage.For example,they allparseddeclarativeandimperativesentences.How-ever, thenatureof thecorpuscanleavemajorgapsin coverage;in thiscase,themanualcontainednoin-terrogatives.

The XLE platform requiresthat a grammarde-clareall the featuresit usesandtheir possibleval-ues.Partof theUrdu featuretableis shownin (16)(thenotationhasbeensimplifiedfor expositorypur-poses).As seenin (16) for QUANT, attributeswhichtakeotherattributesastheirvaluesmustalsobede-clared.An exampleof sucha featurewas seenin(3b) for SPEC which takesQUANT and POSS fea-tures,amongothers,asits values.

(16) PRON-TYPE: � perspossnull � .PROPER: � datelocationnametitle � .PSEM: � locationaldirectional � .PTYPE: � semnosem� .QUANT: � PRED QUANT-TYPE

QUANT-FORM � .The featuredeclarationsof all of the languagesarecomparedfeatureby featureto ensureparallelism.The mostobvioususeof this is to ensurethat thegrammarsencodethesamefeaturesin thesameway.Forexample,atabasiclevel,onefeaturedeclarationmight havespecifiedGEN for genderwhile theoth-ershadchosenthe nameGEND; this divergenceinnamingis regularized.More interestingcasesarise

whenonelanguageusesa featureandanotherdoesnotfor analyzingthesamephenomena.Whenthisisnoticedvia thefeature-tablecomparison,it is deter-minedwhy onegrammarneedsthe featureandtheotherdoesnot,andthusit maybepossibleto elim-inatethefeaturein onegrammaror to addit to an-other.

On a deeperlevel, thefeaturecomparisonis use-ful for conductinga surveyof what constructionseachgrammarhasandhow they areimplemented.For example,if a languagedoesnot havean ADE-GREE (adjectivedegree)feature,the questionwillariseasto whetherthe grammaranalyzescompar-ativeandsuperlativeadjectives.If theydonot, thentheyshouldbeaddedandshouldusetheADEGREE

feature;if theydo,thenthequestionarisesasto whytheydonothavethisfeatureaspartof theiranalysis.

Finally, there is the discussionof problematicconstructions.Thesemay be constructionsthat al-readyhaveanalyseswhichhadbeenagreeduponinthepastbutwhicharenotworkingproperlynowthatmore data has beenconsidered.More frequently,theyarenewconstructionsthatoneof thegrammarsis consideringadding.Possibleanalysesfor thecon-structionare discussedand then one of the gram-marswill incorporatetheanalysisto seewhetheritworks. If the analysisworks, then the othergram-mars will incorporatethe analysis.Constructionsthat have beendiscussedin past ParGrammeet-ingsincludepredicativeadjectives,quantifiers,par-titives, andclefts. Even if not all of the languageshavethe constructionin question,aswas the casewith clefts, the grammarwriters for that languagemay have interestingideason how to analyzeit.Thesegroup discussionshave provenparticularlyusefulin extendinggrammarcoveragein a parallelfashion.

Oncea consensusis reached,it is the responsi-bility of eachgrammarto makesurethat its anal-ysesmatchthe new standard.As such,after eachmeeting,thegrammarwriterswill renamefeatures,changeanalyses,andimplementnewconstructionsinto theirgrammars.Mostof thebasicworkhasnowbeenaccomplished.However, asthegrammarsex-pandcoverage,moreconstructionsneedto beinte-gratedinto the grammars,and theseconstructionstendto beonesfor whichthereis nostandardanaly-sisin thelinguisticliterature;so,differencescaneas-ily arisein theseareas.

6

Page 11: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

4 Conclusion

The experiencesof the ParGramgrammarwritershasshownthattheparallelismof analysisandimple-mentationin theParGramprojectaidsfurthergram-mar developmentefforts. Many of the basicdeci-sions about analysesand formalism have alreadybeenmadein theproject.Thus,thegrammarwriterfor a new languagecanuseexistingtechnologytobootstrapa grammarfor thenewlanguageandcanparseequivalentconstructionsin the existing lan-guagesto seehow to analyzea construction.Thisallows the grammarwriter to focuson morediffi-cultconstructionsnotyetencounteredin theexistinggrammars.

Considerfirst the Japanesegrammarwhich wasstartedin thebeginningof 2001.At theinitial stage,thework of grammardevelopmentinvolvedimple-mentingthebasicconstructionsalreadyanalyzedintheothergrammars.It wasfoundthat thegrammarwriting techniquesandguidelinesto maintainpar-allelismsharedin theParGramprojectcouldbeef-ficiently appliedto the Japanesegrammar. Duringthenextstage,LFGrulesneededfor grammaticalis-suesspecificto Japanesehavebeengraduallyincor-porated,andatthesametime,thebiannualParGrammeetingshavehelpedsignificantlytokeepthegram-marsparallel.Giventhissystem,in ayearandahalf,usingtwo grammarwriters, the Japanesegrammarhasattainedcoverageof 99%for 500sentencesof acopiermanualand95%for 10,000sentencesof aneCRM(Voice-of-Customer)corpus.

Next consider the Norwegian grammarwhichjoinedtheParGramgroupin 1999andalsoempha-sizedslightly differentgoalsfrom theothergroups.Ratherthanprioritizing largetextualcoveragefromtheoutset,theNorwegiangroupgavepriority to thedevelopmentof a coregrammarcoveringall majorconstructiontypesin a principledway basedon theproposalsin (Bresnan,2001)andtheinclusionof asemanticprojectionin additionto thef-structure.Inaddition,timewasspentonimprovingexistinglexi-calresources( � 80,000lemmas)andadaptingthemto theXLE format.Roughlytwoman-yearshasbeenspenton the grammaritself. The ParGramcooper-ationon parallelismhasensuredthat thederivedf-structuresareinterestingin a multilingual context,andthegrammarwill nowserveasabasisfor gram-mardevelopmentin othercloselyrelatedScandina-vian languages.

Thus, the ParGramproject hasshownthat it ispossibleto useasinglegrammardevelopmentplat-

form andaunifiedmethodologyof grammarwritingto developlarge-scalegrammarsfor typologicallydifferentlanguages.Thegrammars’analysesshowalargedegreeof parallelism,despitebeingdevelopedatdifferentsites.Thisis achievedby intensivemeet-ingstwiceayear. Theparallelismcanbeexploitedinapplicationsusingthegrammars:the fewerthedif-ferences,thesimplera multilingual applicationcanbe(see(Frank,1999)on amachine-translationpro-totypeusingParGram).

ReferencesJoanBresnan. 2001. Lexical-FunctionalSyntax.

Blackwell.Miriam Butt, TracyHollowayKing, Marıa-Eugenia

Nino,andFrederiqueSegond.1999.A GrammarWriter’s Cookbook. CSLI Publications.

AnnCopestake.2002. ImplementingTypedFeatureStructure Grammars. CSLI Publications.

BertholdCrysmann,AnetteFrank,BerndKeifer, St.Muller, GunterNeumann,JakubPiskorski,UlrichSchafer, Melanie Siegel,HansUszkoreit,FeiyuXu, Markus Becker, and Hans-Ulrich Krieger.2002.An integratedarchitecturefor shallowanddeepparsing.In Proceedingsof theAnnualMeet-ing of theAssociationfor ComputationalLinguis-tics,Universityof Pennsylvania.

AnetteFrank. 1999. Fromparallelgrammardevel-opmenttowardsmachinetranslation.In Proceed-ingsof MT SummitVII, pages134–142.

JohnT. Maxwell, III andRon Kaplan. 1993. Theinterface betweenphrasal and functional con-straints.ComputationalLingusitics, 19:571–589.

StefanRiezler, Tracy Holloway King, RonaldKa-plan, Dick Crouch, John T. Maxwell, III, andMarkJohnson.2002.Parsingthewall streetjour-nal using a lexical-functionalgrammarand dis-criminative estimationtechniques. In Proceed-ingsof theAnnualMeetingof theAssociationforComputationalLinguistics,University of Penn-sylvania.

PaulSchmidt,Sibylle Rieder, Axel Theofilidis,andThierry Declerck. 1996. Lean formalisms,lin-guistic theory, andapplications:Grammardevel-opmentin alep. In Proceedingsof COLING.

Wolfgang Wahlster, editor. 2000. Verbmobil:Foundations of Speech-to-SpeechTranslation.Springer.

7

Page 12: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

The Grammar Matrix: An Open-SourceStarter-Kit for the RapidDevelopmentof Cross-LinguisticallyConsistentBroad-CoveragePrecision

Grammars

Emily M. Bender and Dan Flickinger and StephanOepenCenterfor theStudyof LanguageandInformation

StanfordUniversity�bender � dan � oe � @csli.stanford.edu

AbstractThe grammarmatrix is an open-source

starter-kit for the development of broad-coverageHPSGs. By using a type hierar-chy to representcross-linguisticgeneraliza-tionsandproviding compatibilitywith otheropen-sourcetoolsfor grammarengineering,evaluation,parsingandgeneration,it facil-itatesnot only quick start-upbut alsorapidgrowth towardsthewidecoveragenecessaryfor robust natural languageprocessingandthe precisionparsesandsemanticrepresen-tationsnecessaryfor naturallanguageunder-standing.

1 Intr oduction

The past decadehas seen the development ofwide-coverageimplementedgrammarsrepresent-ing deeplinguistic analysisof several languagesin several frameworks, including Head-DrivenPhrase Structure Grammar (HPSG), Lexical-FunctionalGrammar(LFG), andLexicalizedTreeAdjoiningGrammar(LTAG). In HPSG, themostex-tensive grammarsarethoseof English(Flickinger,2000), German (Muller & Kasper, 2000), andJapanese(Siegel, 2000; Siegel & Bender, 2002).Despitebeingcouchedin thesamegeneralframe-work and in some casesbeing written in thesameformalismandconsequentlybeingcompati-blewith thesameparsingandgenerationsoftware,thesegrammarsweredevelopedmoreor lessinde-pendentlyof eachother. They eachrepresentbe-tween5 and 15 personyearsof researchefforts,and comprise35–70,000lines of code. Unfor-tunately, most of that researchis undocumentedand the accumulatedanalyses,bestpracticesforgrammarengineering,and tricks of the tradeareonly available through painstakinginspectionofthe grammarsand/or consultationwith their au-thors. This lack of documentationholds acrossframeworks, with certainnotableexceptions,in-

cluding Alshawi (1992),Muller (1999),andButt,King, Nino,& Segond(1999).

Grammarswhich have beenunderdevelopmentfor many yearstendto beverydifficult to mineforinformation,asthey containlayersuponlayersofinteractinganalysesanddecisionsmadein light ofvariousintermediatestagesof thegrammar. As aresult,whenembarkingon the creationof a newgrammarfor anotherlanguage,it seemsalmosteasierto startfromscratchthanto try tomodelit onan existing grammar. This is unfortunate—beingableto leveragethe knowledgeandinfrastructureembeddedin existing grammarswould greatlyac-celeratetheprocessof developingnew ones.At thesametime, thesegrammarsrepresentanuntappedresourcefor thebottom-upexplorationof languageuniversals.

As partof theLinGO consortium’smulti-lingualgrammarengineeringeffort, we aredeveloping a‘grammarmatrix’ or starter-kit, distilling thewis-domof existing grammarsandcodifying anddoc-umentingit in a form thatcanbeusedasthebasisfor new grammars.

In the following sections,we outline the inven-tory of a first, preliminaryversionof thegrammarmatrix, discussthe interactionof basicconstruc-tion typesandsemanticcompositionin unificationgrammarsby meansof a detailedexample, andconsiderextensionsto the core inventory that weforeseeandanevaluationmethodologyfor thema-trix proper.

2 Preliminary Developmentof MatrixWe have produceda preliminary version of thegrammarmatrix relying heavily on the LinGOproject’s English ResourceGrammar, and to alesserextent on the Japanesegrammardevelopedjointly betweenDFKI Saarbrucken(Germany) andYY Technologies(MountainView, CA). Thisearlyversionof thematrixcomprisesthefollowing com-

8

Page 13: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ponents:�Typesdefiningthebasicfeaturegeometryandtechnicaldevices(e.g.,for list manipulation).�Typesassociatedwith Minimal RecursionSe-mantics(see,e.g.,Copestake, Lascarides,&Flickinger, 2001), a meaningrepresentationlanguagewhich hasbeenshown to be well-suitedfor semanticcompositionin typedfea-ture structuregrammars.This portion of thegrammarmatrix includesa hierarchyof rela-tion types,typesandconstraintsfor theprop-agationof semanticinformation throughthephrasestructuretree,a representationof illo-cutionaryforce, andprovisions for grammarruleswhichmake semanticcontributions.�Generalclassesof rules, including deriva-tional and inflectional (lexical) rules, unaryandbinaryphrasestructurerules,headedandnon-headedrules,andhead-initialandhead-final rules. Theserule classesinclude im-plementationsof generalprinciplesof HPSG,like, for example,theHeadFeatureandNon-Local FeaturePrinciples.�Typesfor basicconstructionssuchas head-complement, head-specifier, head-subject,head-filler, and head-modifierrules, coordi-nation, as well as more specializedclassesof constructions,suchasrelative clausesandnoun-nouncompounding.Unlike in specificgrammars,thesetypesdo not imposeany or-deringon theirdaughtersin thegrammarma-trix.

Includedwith the matrix areconfigurationandparameterfiles for the LKB grammarengineeringenvironment(Copestake,2002).

Although small, this preliminary version ofthe matrix already reflects the main goals ofthe project: (i) Consistentwith other work inHPSG, semanticrepresentationsand in particularthesyntax-semanticsinterfacearedevelopedin de-tail; (ii) the typesof thematrix areeachrepresen-tationsof generalizationsacrosslinguistic objectsandacrosslanguages;and(iii) the richnessof thematrixandtheincorporationof fileswhichconnectit with the LKB allow for extremelyquick start-upasthematrix is appliedto new languages.

SinceFebruary2002,thispreliminaryversionofthematrix hasbeenin useat two Norwegianuni-versities,one working towardsa broad-coverage

referenceimplementationof Norwegian (NTNU),theother—for thetimebeing—focusedonspecificaspectsof clausestructureandlexical description(Oslo University). In the first experimentwiththe matrix, at NTNU, basicNorwegian sentenceswereparsingandproducingreasonablesemanticswithin two hoursof downloadingthematrix files.Linguistic coverageshouldscaleup quickly, sincethe foundationsuppliedby thematrix is designednotonly to provideaquickstart,but alsotosupportlong-term developmentof broad-coveragegram-mars.Both initiativeshave confirmedtheutility ofthematrix starterkit andalreadyhave contributedto a seriesof discussionson cross-lingualHPSG

designaspects,specifically in the areasof argu-ment structurerepresentationsin the lexicon andbasicassumptionsaboutconstituentstructure(inoneview, Norwegian exhibits a VSO topologyinthemainclause).Theusergroupshave suggestedrefinementsandextensionsof thebasicinventory,andit is expectedthatgeneralsolutions,asthey areidentified jointly, will propagateinto the existinggrammarstoo.

3 A DetailedExampleAs an exampleof the level of detail involved inthe grammarmatrix, in this sectionwe considerthe analysisof intersective and scopalmodifica-tion. Thematrixis built to giveMinimal RecursionSemantics(MRS; Copestake et al., 2001; Copes-take,Flickinger, Sag,& Pollard,1999;Copestake,Flickinger, Malouf, Riehemann,& Sag,1995)rep-resentations. The two English examplesin (1)exemplify thedifferencebetweenintersective andscopalmodification:1

(1) a. KeanustudiedKungFu onaspaceship.

b. KeanuprobablystudiedKungFu.

The MRSs for (1a-b) (abstractingaway fromagreementinformation) are given in (2) and (3).The MRSsareorderedtuplesconsistingof a tophandle( ��� in bothcases),aninstanceoreventvari-able( � in bothcases),abagof elementarypredica-tions(eps),andabagof scopeconstraints(in thesecases,QEQ constraintsor ‘equal modulo quanti-fiers’). In a well-formedMRS, thehandlescanbe

1Theseexamplesalso differ in that probably is a pre-headmodifier while on a spaceshipis a post-headmodifier.This word-orderdistinction cross-cutsthe semanticdistinc-tion, andour focusis on the latter, sowe won’t considertheword-orderaspectsof theseexampleshere.

9

Page 14: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

identifiedin oneor morewaysrespectingthescopeconstraintssuchthatthedependenciesbetweentheepsform atree.For adetaileddescriptionof MRS,seetheworkscitedabove. Here,we will focusonthedifferencebetweentheintersective modifieron(a spaceship) andthescopalmodifierprobably.

In (2), theepcontributedby on (‘on-rel’) sharesits handle( � � ) with theepcontributedby theverbit is modifying (‘study-rel’). As such,thetwo willalwayshave thesamescope;no quantifiercanin-tervene.Further, thesecondargumentof theon-rel( � ) is theeventvariableof thestudy-rel. Thefirstargument,�"! , is theeventvariableof theon-relandthethird argument,# , is theinstancevariableof thespaceship-rel.

(2) $ h1,e,%h1:prpstn-rel(h2),h3:def-np-rel(x,h4,h5),

h6:named-rel(x,‘K eanu’),h7:study-rel(e,x, y),h8:def-np-rel(y, h9,h10),h11:named-rel(y, ‘K ungFu’), h7:on-rel(e& , e,z),h12:a-quant-rel(z,h13,h14),h15:spaceship-rel(z)' ,%

h2 QEQ h7,h4 QEQ h6,h19QEQ h11,h13QEQ h15 ')(

In (3), theepcontributedby thescopalmodifierprobably (‘probably-rel’) hasits own handle( � � )which is not sharedby anything. Furthermore,ittakesa handle( �+* ) ratherthanthe event variableof thestudy-relasits argument. � * is equalmod-ulo quantifiers(QEQ) to thehandleof thestudy-rel( �+, ), and �-� is equalmoduloquantifiersto thear-gumentof theprpstn-rel( �+. ). Theprpstn-relis theeprepresentingtheillocutionaryforceof thewholeexpression.This meansthatquantifiersassociatedwith theNPsKeanuandKungFu canscopeinsideor outsideprobably.

(3) $ h1,e,%h1:prpstn-rel(h2),h3:def-np-rel(x,h4,h5),

h6:named-rel(x,‘K eanu’),h7:probably-rel(h8), h9:study-rel(e,x, y),h10:def-np-rel(y, h11,h12),h13:named-rel(y, ‘K ungFu’) ' ,%

h2 QEQ h7,h4 QEQ h6,h8 QEQ h9,h11QEQ h13 ')(

While thedetailsof modifier placement,whichpartsof speechcanmodify whichkindsof phrases,etc., differ acrosslanguages,we believe that alllanguagesdisplayadistinctionbetweenscopalandintersective modification. Accordingly, the types

isect-mod-phrase := head-mod-phr-simple &[ HEAD-DTR.SYNSEM.LOCAL

[ CONT [ TOP #hand,INDEX #index ],KEYS.MESSAGE 0-dlist ],

NON-HEAD-DTR.SYNSEM.LOCAL[ CAT.HEAD.MOD <[ LOCAL isect-mod ]>,

CONT.TOP #hand ],C-CONT.INDEX #index ].

Figure1: TDL descriptionof isect-mod-phrase

scopal-mod-phrase := head-mod-phr-simple &[ NON-HEAD-DTR.SYNSEM.LOCAL

[ CAT.HEAD.MOD <[ LOCAL scopal-mod ]>,CONT.INDEX #index ],

C-CONT.INDEX #index ].

Figure2: TDL descriptionof scopal-mod-phrase

necessaryfor describingthesetwo kindsof modi-ficationareincludedin thematrix.

The types isect-mod-phrase and scopal-mod-phrase(shown in Figures1 and2) encodethe in-formationnecessaryto build up in acompositionalmannerthe modifier portionsof the MRSs in (2)and(3).

Thesetypes are embeddedin the type hierar-chy of thematrix. Throughtheir supertypehead-mod-phr-simplethey inherit informationcommonto many typesof phrases,including thebasicfea-ture geometry, headfeatureandnon-localfeaturepassing,and semanticcompositionality. Thesetypesalsohave subtypesin the matrix specifyingthetwo word-orderpossibilities(pre-or post-headmodifiers),giving a total of four subtypes.2

The most important differencebetweenthesetypesis in the treatmentof thehandleof theheaddaughter’s semantics,to distinguish intersectiveandscopalmodification. In isect-mod-phrase, thetop handles(TOP)of theheadandnon-head(i.e.,modifier) daughtersareidentified(#hand). Thisallowsfor MRSslike(2) wheretheepscontributedby thehead(‘study-rel’) andthemodifier(‘on-rel’)take thesamescope.Thetypescopal-mod-phrasebearsno suchconstraint. This allows for MRSslike (3) wherethemodifier’ssemanticcontribution(‘probably-rel’) takesthehandleof the head’s se-mantics(‘study-rel’) as its argument,so that themodifieroutscopesthehead.In bothtypesof mod-

2All four subtypesare provided on the theory that mostlanguageswill make useof all or mostof them.

10

Page 15: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ifier phrase,a constraintinheritedfrom thesuper-typeensuresthatthehandleof themodifieris alsothehandleof thewholephrase.

The constraintson the LOCAL value insidethe modifier’s MOD value regulate which lexi-cal items can appearin which kind of phrase.Intersective modifiers specify lexically that theyare [ MOD / [ LOCAL isect-mod] 0 ] and sco-pal modifiers specify lexically that they are[ MOD / [ LOCAL scopal-mod] 0 ].3 Thesecon-straintsexemplify thekind of informationthatwillbedevelopedin thelexical hierarchyof thematrix.

It is characteristicof broad-coveragegrammarsthat every particularanalysisinteractswith manyotheranalyses.Modularizationis anon-goingcon-cern,both for maintainabilityof individual gram-mars,andfor providing the right level of abstrac-tion in thematrix. For thesamereasons,we haveonly beenableto touchon thehighlightsof these-manticanalysisof modificationhere,but hopethatthis quick tour will suffice to illustrate the extentof the jump-startthematrix cangive in thedevel-opmentof new grammars.

4 Future Extensions

Theinitial versionof thematrix,while sufficient tosupportsomeuseful grammarwork, will requiresubstantialfurther developmenton several fronts,including lexical representation,syntacticgener-alization, sociolinguisticvariation, processingis-sues,andevaluation.This first versiondrew mostheavily from the implementationof the Englishgrammar, with somefurther insightsdrawn fromthe grammarof Japanese.Extensionsto the ma-trix will bebasedon carefulstudyof existing im-plementedgrammarsfor otherlanguages,notablyGerman,Spanishand Japanese,as well as feed-backfrom thoseusingthefirst versionof thema-trix.

For lexical representation,one of the most ur-gent needsis to provide a language-independenttype hierarchyfor the lexicon, at leastfor majorpartsof speech,establishingthemechanismsusedfor linking syntacticsubcategorization to seman-tic predicate-argumentstructure.Lexical rulespro-vide a secondmechanismfor expressinggeneral-

3Notethatthereareno furthersubtypesof LOCAL valuesbeyond isect-modandscopal-mod. Sincethesegrammarsdonot make extensive useof subtypesof LOCAL values,theywereavailablefor encodingthisdistinction.Alternativesolu-tionsincludepositinga new feature.

izationswithin thelexicon,andoffer readyoppor-tunities for cross-linguisticabstractionsfor bothinflectionalandderivational regularities. Work isalso progressingon establishinga standardrela-tional database(usingPostgreSQL)for storingin-formation for the lexical entriesthemselves, im-proving both scalability and clarity comparedtothe currentsimple text file representation.Form-basedtoolswill beprovidedboth for constructinglexical entriesandfor viewing thecontentsof thelexicon.

Theprimaryfocusof work onsyntacticgeneral-ization in the matrix is to supportmore freedomin word order, for both complementsand modi-fiers. The first stepwill be a relatively conserva-tive extensionalongthelinesof Netter(1996),al-lowing thegrammarwriter morecontrolover howa headcombineswith complementsof differenttypes,andtheir interleaving with modifierphrases.Otherareasof immediatecross-linguisticinterestinclude the hierarchyof headtypes,control phe-nomena,clitics, auxiliary verbs,noun-nouncom-pounds,and more generally, phenomenathat in-volve theword/phrasedistinction,suchasnounin-corporation.A studyof theexisting grammarsforEnglish, German,Japanese,and Spanishrevealsa high degreeof language-specificityfor severalof thesephenomena,but alsosuggestspromiseofreusableabstractions.

Severalkindsof sociolinguisticvariationrequireextensionsto thematrix, includinggrammaticizedaspectsof pragmaticssuchaspolitenessandem-pathy, aswell asdialectandregisteralternations.Thegrammarof Japaneseprovidesastartingpointfor representationsof bothempathyandpoliteness.Implementationsof familiar vs. formal verbformsin GermanandSpanishprovide further instancesof politenessto help build thecross-linguisticab-stractions. Extensionsfor dialect variation willbuild on someexploratory work in adaptingtheEnglish grammarto support American, British,andAustralianregionalisms,bothlexical andsyn-tactic,while restrictingdialectmixture in genera-tion andassociatedspuriousambiguityin parsing.

While the developmentof the matrix will bebuilt largelyontheLKB platform,supportwill alsobe neededfor using the emerging grammarsonotherprocessingplatforms,andfor linking to otherpackagesfor pre-processingthe linguistic input.Severalotherplatformsexist which canefficientlyparsetext using the existing grammars,includ-

11

Page 16: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ing the PET systemdevelopedin C121 at SaarlandUniversity (Germany) and the DFKI (Callmeier,2000); the PAGE systemdevelopedin Lisp at theDFKI (Uszkoreit et al., 1994);theLiLFeS systemdevelopedat Tokyo University (Makino, Yoshida,Torisawa, & Tsujii, 1998),anda parallelprocess-ing systemdevelopedin Objective C at Delft Uni-versity (The Netherlands;vanLohuizen, 2002).As part of the matrix package,sampleconfigura-tion files anddocumentationwill be provided forat leastsomeof theseadditionalplatforms.

Existing pre-processingpackagescanalsosig-nificantly reducethe effort required to developa new grammar, particularly for coping with themorphology/syntaxinterface. For example, theChaSenpackagefor segmenting Japaneseinputinto words and morphemes(Asahara & Mat-sumoto,2000)hasbeenlinked to at leastthe LKB

and PET systems. Support for connectingim-plementationsof language-specificpre-processingpackagesof this kind will be preserved and ex-tendedas the matrix develops. Likewise, config-urationfiles areincludedto supportgeneration,atleastwithin the LKB, provided that the grammarconformsto certain assumptionsabout semanticrepresentationusing the Minimal RecursionSe-manticsframework.

Finally, amethodologyis underdevelopmentforconstructingandusingtestsuitesorganizedarounda typologyof linguistic phenomena,usingthe im-plementationplatform of the [incr tsdb()] profil-ing package(Oepen& Flickinger, 1998; Oepen& Callmeier, 2000). Thesetestsuiteswill enablebettercommunicationaboutcurrentcoverageof agivengrammarbuilt usingthematrix,andserveasthebasisfor identifyingadditionalphenomenathatneedto beaddressedcross-linguisticallywithin thematrix. Of course,the developmentof the typol-ogy of phenomenais itself a major undertakingfor which a systematiccross-linguisticapproachwill be needed,a discussionof which is outsidethe scopeof this report. But the intent is to seedthis classificationschemewith a set of relativelycoarse-grainedphenomenonclassesdrawn fromtheexisting grammars,thenrefinethetypologyasit is appliedto theseandnew grammarsbuilt usingthematrix.

5 CaseStudies

Oneimportantpartof thematrixpackagewill bealibrary of phenomenon-basedanalysesdrawn from

theexistinggrammarsandover time from usersofthe matrix, to provide working examplesof howthematrixcanbeappliedandextended.Eachcasestudywill bea setof grammarfiles, simplifiedforrelevance,alongwith documentationof the anal-ysis, and a test suite of samplesentenceswhichdefinethe rangeof datacoveredby the analysis.This library, too, will beorganizedaroundthe ty-pology of phenomenaintroducedabove, but willalsomake explicit referenceto languagefamilies,sinceboth similarities anddifferencesamongre-lated languageswill be of interestin thesecasestudies. Examplesto be includedin the first re-leaseof this library includenumeralclassifiersinJapanese,subjectpro drop in Spanish,partial-VPfronting in German,andverbdiathesisin Norwe-gian.

6 Evaluation and Evolution

The matrix itself is not a grammarbut a collec-tion of generalizationsacrossgrammars.As such,it cannotbetesteddirectly oncorporafrom partic-ular languages,andwe mustfind othermeansofevaluation. We envision overall evaluationof thematrix basedon casestudiesof its performancein helping grammarengineersquickly start newgrammarsand in helping themscalethosegram-marsup. Evaluation in detail will basedon au-tomatabledeletion/substitution metrics,i.e., toolsthat determinewhich types from the matrix getusedasis, which getusedwith modifications,andwhichgetignoredin variousmatrix-derivedgram-mars.Furthermore,if thematrixevolvesto includedefeasibleconstraints,thesetoolswill checkwhichconstraintsget overriddenand whetherthe valuechosenis indeedcommonenoughto bemotivatedasa default value.This evaluationin detailshouldbe pairedwith feedbackfrom the grammarengi-neersto determinewhy changesweremade.

Themaingoalof evaluationis, of course,to im-prove the matrix over time. This raisesthe ques-tion of how to propagatechangesin thematrix togrammarsbasedonearlierversions.Thefollowingthreestrategies(meantto beusedin combination)seempromising:(i) segregatechangesthatareim-portantto sync to (e.g.,changesthat affect MRSoutputs,fundamentalchangesto importantanal-yses),(ii) develop a methodologyfor communi-catingchangesin thematrix, their motivationandtheir implementationto the usercommunity, and(iii) developtoolsfor semi-automatingresynching

12

Page 17: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

of existing grammarsto upgradesof the matrix.Thesetoolscouldusethetypehierarchyto predictwhereconflictsarelikely to ariseandbring theseto theengineer’sattention,possiblyinspiredby theapproachunderdevelopmentat CSLI for the dy-namicmaintenanceof theLinGO Redwoodstree-bank(Oepenetal., 2002).

Finally, while initial developmentof the ma-trix hasbeenandwill continueto be highly cen-tralized,we hopeto provide supportfor proposedmatrix improvementsfrom the user community.User feedbackwill alreadycomein the form ofcasestudiesfor the library as discussedin Sec-tion 5 above, but alsopotentially in proposalsformodificationof thematrixdrawing onexperiencesin grammardevelopment. In order to provideuserswith somecross-linguisticcontext in whichto developandevaluatesuchproposalsthemselves,we intendto provide somesamplematrix-derivedgrammarsand correspondingtestsuiteswith thematrix. A usercouldthusmakeaproposedchangeto the matrix, run the testsuitesfor several lan-guagesusing the suppliedgrammarswhich drawfrom that changedmatrix, and use [incr tsdb()]to determinewhichphenomenahavebeenaffectedby the change. It is clear that full automationofthisevaluationprocesswill bedifficult, but at leastsomeclassesof changesto the matrix will per-mit this kind of quick cross-linguisticfeedbacktouserswith only a modestamountof additionalin-frastructure.

7 Conclusion

This projectcarrieslinguistic, computational,andpracticalinterest.Thelinguistic interestlies in theHPSG community’s generalbottom-up approachto languageuniversals,which involvesaiming forgoodcoverageof a variety of languagesfirst, andleaving the taskof what they have in commonforlater. (Of course,theorybuilding is never purelydata-driven, and therearesubstantive hypotheseswithin HPSG about languageuniversals.) Nowthatwehave implementationswith fairly extensivecoveragefor a somewhat typologicallydiversesetof languages,it is agoodtimeto take thenext stepin this program,working to extractandgeneralizewhatis similaracrosstheseexistingwide-coveragegrammars.Moreover, the centralrole of typesintherepresentationof linguistic generalizationsen-ablesthekindof underspecificationwhichis usefulfor expressingwhatis commonamongrelatedlan-

guageswhile allowing for the further specializa-tion which necessarilydistinguishesonelanguagefrom another.

The computationalinterestis threefold. Firstthere is the questionof what formal devices thegrammarmatrix will require. Should it includedefaults?Whataboutdomainunion(linearizationtheory)? The selectionanddeploymentof formaldevicesshouldbe informedby on-goingresearchon processingschemes,andherethecrosslinguis-tic perspective canbeparticularlyhelpful. Wherethereareseveral equivalent analysesof the samelinguistic phenomena(e.g., morphosyntacticam-biguity or optionality), the choiceof analysiscanhave processingimplicationsthat aren’t necessar-ily apparentin a singlegrammar. Second,havingasetof wide-coverageHPSGswith fairly standard-ized fundamentalscould prove interestingfor re-searchon stochasticprocessingand disambigua-tion, especiallyif the languagesdiffer in grossty-pological featuressuch as word order. Finally,there are also computationalissuesinvolved inhow the grammarmatrix would evolve over timeas it is usedin new grammars. The matrix en-ablesthe developerof a grammarfor a new lan-guageto get a quick starton producinga systemthat parsesandgenerateswith non-trivial seman-tics,while alsobuilding thefoundationfor awide-coveragegrammarof thelanguage.But thematrixitself may well changein parallelwith the devel-opmentof the grammarfor a particularlanguage,so appropriatemechanismsmustbe developedtosupportthemerging of enhancementsto both.

Thereis alsopracticalindustrialbenefitto thisproject. Companiesthat are consumersof thesegrammarsbenefitwhengrammarsof multiple lan-guageswork with thesameparsingandgenerationalgorithmsandproducestandardizedsemanticrep-resentationsderivedfrom arich, linguisticallymo-tivatedsyntax-semanticsinterface. More impor-tantly, thegrammarmatrixwill helpto removeoneof theprimaryremainingobstaclesto commercialdeploymentof grammarsof thistypeandindeedofthecommercialuseof deeplinguisticanalysis:theimmensecostof developingtheresource.

Acknowledgements

Since the grammar matrix draws on prior re-searchand existing grammars,it necessarilyre-flects contributions from many people. RobMalouf, Jeff Smith, John Beavers, and Kathryn

13

Page 18: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Campbell-Kiblerhave contributed to the LinGOERG; MelanieSiegel is the original developerfortheJapanesegrammar. Tim Baldwin,Ann Copes-take, Ivan Sag,Tom Wasow, and other membersof theLinGO LaboratoryatCSLI havehadagreatdealof influenceon thedesignof thegrammaticalanalysesandcorrespondingMRS representations.Warmestthanksto LarsHellanandhis colleaguesat NTNU andJanTore Lønningandhis studentsat OsloUniversity for their cooperation,patience,andtolerance.

References

Alshawi, H. (Ed.). (1992).TheCoreLanguageEngine.Cambridge,MA: MIT Press.

Asahara,M., & Matsumoto,Y. (2000).Extendedmod-els and tools for high-performancepart-of-speechtagger. In Proceedingsof the 18th InternationalConferenceon ComputationalLinguistics(pp. 21–27). Saarbrucken,Germany.

Butt, M., King, T. H., Nino, M.-E., & Segond, F.(1999). A grammar writer’s cookbook. Stanford,CA: CSLI Publications.

Callmeier, U. (2000). PET — A platform for ex-perimentationwith efficient HPSGprocessingtech-niques.Natural Language Engineering, 6 (1) (Spe-cial Issueon EfficientProcessingwith HPSG), 99–108.

Copestake, A. (2002). Implementingtyped featurestructure grammars. Stanford,CA: CSLI Publica-tions.

Copestake, A., Flickinger, D., Malouf, R., Riehemann,S., & Sag,I. (1995). Translationusingminimal re-cursionsemantics.In Proceedingsof the Sixth In-ternationalConferenceon Theoretical andMethod-ological Issuesin Machine Translation. Leuven,Belgium.

Copestake, A., Flickinger, D. P., Sag, I. A., & Pol-lard, C. (1999). Minimal Recursion Semantics.Anintroduction. in preparation,CSLI Stanford,Stan-ford, CA.

Copestake,A., Lascarides,A., & Flickinger, D. (2001).An algebrafor semanticconstructionin constraint-basedgrammars.In Proceedingsof the 39th Meet-ing of theAssociationfor ComputationalLinguistics.Toulouse,France.

Flickinger, D. (2000). On building a more efficientgrammarby exploiting types.Natural LanguageEn-gineering, 6 (1) (SpecialIssueon EfficientProcess-ing with HPSG), 15–28.

vanLohuizen, M. (2002). Efficient and thread-safeunificationwith LinGO. In S.Oepen,D. Flickinger,J. Tsujii, & H. Uszkoreit (Eds.), Collaborativelanguage engineering. A case study in efficient

grammar-basedprocessing. Stanford, CA: CSLIPublications.(forthcoming)

Makino, T., Yoshida,M., Torisawa, K., & Tsujii, J.(1998).LiLFeS— towardsapracticalHPSGparser.In Proceedingsof the17thInternationalConferenceon ComputationalLinguisticsand the 36th AnnualMeetingof the Associationfor ComputationalLin-guistics(pp.807–11). Montreal,Canada.

Muller, S. (1999). Deutsche syntaxdeklarativ. Head-DrivenPhraseStructureGrammarfur dasDeutsche.Tubingen,Germany: Max NiemeyerVerlag.

Muller, S., & Kasper, W. (2000). HPSGanalysisofGerman. In W. Wahlster (Ed.), Verbmobil. Foun-dations of speech-to-speech translation (ArtificialIntelligenceed., pp. 238–253). Berlin, Germany:Springer.

Netter, K. (1996). Functionalcategoriesin an HPSGfor German. Unpublished doctoral dissertation,SaarlandUniversity, Saarbrucken,Germany.

Oepen,S., & Callmeier, U. (2000). Measurefor mea-sure: Parsercross-fertilization.Towards increasedcomponentcomparability and exchange. In Pro-ceedingsof the6th InternationalWorkshopon Pars-ing Technologies(pp.183–194).Trento,Italy.

Oepen,S., & Flickinger, D. P. (1998). Towardssys-tematicgrammarprofiling. Testsuitetechnologytenyearsafter. Journal of ComputerSpeech and Lan-guage, 12 (4) (SpecialIssueon Evaluation), 411–436.

Oepen,S., Toutanova, K., Shieber, S., Manning, C.,Flickinger, D., & Brants, T. (2002). The LinGORedwoodstreebank.Motivationandpreliminaryap-plications. In Proceedingsof the 19th InternationalConferenceon ComputationalLinguistics. Taipei,Taiwan.

Siegel, M. (2000). HPSG analysis of Japanese.In W. Wahlster (Ed.), Verbmobil. Foundationsofspeech-to-speech translation(Artificial Intelligenceed.,pp.265–280). Berlin, Germany: Springer.

Siegel, M., & Bender, E. M. (2002). Efficient deepprocessingof japanese.In Proceedingsof the 19thInternationalConferenceonComputationalLinguis-tics. Taipei,Taiwan.

Uszkoreit, H., Backofen, R., Busemann,S., Diagne,A. K., Hinkelman, E. A., Kasper, W., Kiefer, B.,Krieger, H.-U., Netter, K., Neumann,G., Oepen,S.,& Spackman,S. P. (1994). DISCO — an HPSG-basedNLP systemand its applicationfor appoint-mentscheduling.In Proceedingsof the 15th Inter-national Conferenceon ComputationalLinguistics.Kyoto,Japan.

14

Page 19: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Parallel Distrib uted Grammar Engineering for Practical Applications

StephanOepen3 , Emily M. Bender3 , Uli Callmeier 4 , Dan Flickinger 3)4 , Melanie Siegel53 CSLI Stanford 4 YY Technologies 5 DFKI GmbHStanford(CA) MountainView (CA) Saarbrucken(Germany)67 8 oe

benderdan 9 :; @csli.stanford.edu < uc

dan = @yy.com [email protected]

Abstract

Basedon a detailedcasestudyof paral-lel grammardevelopmentdistributedacrosstwo sites, we review someof the require-mentsfor regressiontestingin grammaren-gineering,summarizeour approachto sys-tematiccompetenceandperformanceprofil-ing, anddiscussour experiencewith gram-mardevelopmentfor a commercialapplica-tion. If possible,theworkshoppresentationwill beorganizedarounda softwaredemon-stration.

1 Background

The production of large-scale constraint-basedgrammarsandsuitableprocessingenvironmentsisa labour- andtime-intensive processthat, maybe,hasbecomesomewhat of a growth industry overthepastfew years,ascompaniesexploreproductsthatincorporategrammar-basedlanguageprocess-ing. Many broad-coveragegrammarshave beendevelopedover severalyears,sometimesdecades,typically coordinatedby a singlegrammarianwhowould often draw on additionalcontributors (e.g.thethreeHPSG implementationsdevelopedaspartof the VerbMobil effort, seeFlickinger, Copes-take, & Sag,2000, Muller & Kasper, 2000, andSiegel, 2000; or the LFG implementationsdevel-opedwithin theParGramconsortium,Butt, King,Nino,& Segond,1999).

More recently, we also find genuinelysharedand distributed development of broad-coveragegrammars,andwewill useonesuchinitiativeasanexample—viz.anopen-sourceHPSG implementa-tion for Japanesejointly developedbetweenDFKISaarbrucken (Germany) and YY Technologies(MountainView, CA)—to demonstratethetechno-logical and methodologicalchallengespresentindistributedgrammarandsystemengineering.

2 Parallel Distrib uted GrammarDevelopment—ACaseStudy

TheJapanesegrammarbuilds on earlierwork per-formed jointly betweenDFKI and the Computa-tional LinguisticsDepartmentat SaarlandUniver-sity (Germany) within VerbMobil; much like fortheGermanVerbMobilgrammar, two peoplewerecontributing to thegrammarin parallel,onebuild-ing out syntacticanalyses,the otherchargedwithintegratingsemanticcompositioninto the syntax.This relatively strict separationof responsibilitiesmostly enabledgrammariansto serialize incre-mentaldevelopmentof the resource:the syntacti-cianwould supplya grammarwith extendedcov-erageto thesemanticistand,at theonsetof thefol-lowing iteration,startsubsequentwork on syntaxfrom therevisedgrammar.

In theDFKI –YY cooperationthesituationwasquite different. Over a period of eight months,both partnershada grammarianworking on syn-tax and semanticssimultaneouslyon a day-to-day basis; both grammarianswere submittingchangesto ajoint, version-controlledsourcerepos-itory andusuallywould start the work day by re-trieving the most recentrevisions. At the sametime, product building and the developmentofso-called‘domainlibraries’ (structuredcollectionsof knowledgeabouta specificdomainthat is in-stantiatedfrom semanticrepresentationsdeliveredfromgrammaticalanalysis)atYY alreadyincorpo-ratedthe grammaranddependedon it for actual,customer-specificcontracts.Due to a continuousdemandfor improvementsin coverageandanaly-sisaccuracy, thegrammarusedin themainproductline would be updatedfrom the currentdevelop-mentversionaboutonceor twice a week.Parallelto work on theJapanesegrammar(andsimultane-ouswork on grammarsfor EnglishandSpanish),both the grammardevelopmentenvironment(the

15

Page 20: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

open-sourceLKB system;Copestake, 2002) andthe HPSG run-time componentpowering the YYlinguistic analysisengine (the open-sourcePET

parser;Callmeier, 2002) continuedto evolve, asdid theYY-proprietarymappingof meaningrepre-sentationsextractedfrom theHPSG grammarsintodomainknowledge—allcentralpartsof acomplexsystemof interactingcomponentsandconstraints.

As has been argued before (see, for exam-ple, Oepen& Flickinger, 1998), the natureof alarge-scaleconstraint-basedgrammarandthesub-tle interactionsof lexical andconstructionalcon-straintsmake it virtually impossibleto predicthowa changein onepartof thegrammaraffectsover-all systembehaviour. A relatively minor repairinone lexical class,numeraladjectives as in ‘ threebookswereordered’for instance,will have thepo-tentialof breakingtheinteractionof thatclasswiththeconstructionderiving named(numeric)entitiesfrom a numeral(e.g. as in ‘ three is my favouritenumber’) or the partitive construction(e.g. as in‘ three have arrived already’). A ripple effect ofa single changecan thus corrupt the semanticsproducedfor any of thesecasesand in the con-sequencecausefailure or incorrectbehaviour inthe back-endsystem. In addition to thesequal-ity assurancerequirementson grammaticalcover-ageandcorrectness,theYY application(like mostapplicationsfor grammar-basedlinguistic analy-sis) utilizes a setof hand-constructedparserank-ing heuristicsthat enablesthe parserto operatein best-firstsearchmodeand to return only onereading,i.e. theanalysisthat is rankedbestby theheuristiccomponent.The parserankingmachin-ery builds on preferencesthat areassociatedwithindividualor classesof lexical itemsandconstruc-tions. Thesetof preferencesis maintainedin par-allel to the grammar, in a senseproviding a layerof performance-orientedannotationsoverthebasicbuilding blocksof the corecompetencegrammar.Withoutdiscussingthedetailsof theparserankingapproach,it createsan additionalelementof un-certaintyin assessinggrammarchanges:sincethepreferencefor a specific analysisresultsimplic-itly from a seriesof local preferences(of lexicalitems and constructionscontributing to the com-pletederivation), introducingadditionalelements(i.e. new localor globalambiguity)into thesearchspaceandsubjectingthemto the partial orderingcanquickly skew theoverall result.

Summingup,thegrammarandapplicationengi-

neeringexamplepresentedhereillustratesa num-ber of highly typical requirementson the engi-neeringenvironment. First, all grammariansandsystemengineersparticipatingin thedevelopmentprocessneedto keepfrequent,detailed,andaccu-raterecordsof a largenumberof relevantparame-ters,includingbut not limited to grammaticalcov-erage,correctnessof syntacticanalysesand cor-respondingsemanticforms, parseselectionaccu-racy, andoverall systemperformance.Second,asmodificationsto the systemasa whole aremadedaily—andsometimesseveraltimeseachday—alldevelopersmust be able to assessthe impact ofrecentchangesandtrack their effectson all rele-vantparameters;gatheringthedataand analyzingit mustbesimple,fast,andautomatedasmuchaspossible.Third, notall modifications(to thegram-mar or underlyingsoftware)will result in ‘mono-tonic’ or backwards-compatibleeffects. A changein thetreatmentof optionalnominalcomplements,for example, may affect virtually all derivationtreesand rendera comparisonof resultsat thislevel uninformative. At thesametime,a primarilysyntacticchangeof thisnaturewill notcauseanef-fect in associatedmeaningrepresentations,sothata semanticequivalencetest over analysesshouldbe expectedto yield an exact matchto earlierre-sults.Hence,themachineryfor representationandcomparisonof relevant parametersneedsto facil-itate user-level specificationof informative testsandevolution criteria. Finally, themetricsusedintrackinggrammardevelopmentcannotbeisolatedfrom measurementsof systemresourceconsump-tion and overall performance(specificpropertiesof a grammarmay trigger idiosyncrasiesor soft-ware bugs in a particularversionof the process-ing system);therefore,andto enableexchangeofreferencepointsandcomparabilityof experiments,grammariansandsystemdevelopersalike shouldusethesame,homogenuoussetof relevantparam-eters.

3 Integrated CompetenceandPerformanceProfiling

The integratedcompetenceandperformancepro-filing methodology and associatedengineeringplatform,dubbed[incr tsdb()] (Oepen& Callmeier,2000)1 andreviewed in the remainderof this sec-

1See ‘http://www.coli.uni-sb.de/itsdb/ ’for the (draft) [incr tsdb()] usermanual,pronunciationrules,andinstructionsonobtainingandinstallingthepackage.

16

Page 21: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

tion, wasdesignedto meetall of therequirementsidentifiedin theDFKI –YY casestudy. Generallyspeaking,the [incr tsdb()] environment is an in-tegratedpackagefor diagnostics,evaluation,andbenchmarkingin practical grammarand systemengineering.The toolkit implementsan approachto grammardevelopmentandsystemoptimizationthat builds on preciseempiricaldataandsystem-atic experimentation,asit hasbeenadvocatedby,amongothers,Erbach& Uszkoreit (1990),Erbach(1991),andCarroll (1994). [incr tsdb()] hasbeenintegratedwith, as of June2002, nine differentconstraint-basedgrammardevelopmentandpars-ing systems(includingbothenvironmentsin useatYY, i.e. the LKB and PET), thusproviding a pre-standardreferencepoint for a relatively large(andgrowing) communityof NLP developers.The[incrtsdb()] environmentbuilds on the following com-ponentsandmodules:> test and referencedata storedwith annota-

tions in a structureddatabase;annotationscanrangefrom minimal information(uniquetest item identifier, item origin, lengthet al.)to fine-grainedlinguistic classifications(e.g.regardinggrammaticalityandlinguistic phe-nomenapresentedin anitem),asthey arerep-resentedin theTSNLP testsuites,for example(Oepen,Netter, & Klein, 1997);> tools to browse the available data, identifysuitablesubsetsand feed them through theanalysis componentof processingsystemslike the LKB and PET, LiLFeS (Makino,Yoshida,Torisawa, & Tsujii, 1998), TRALE

(Penn,2000),PAGE (Uszkoreit et al., 1994),andothers;> theability to gatheramultitudeof preciseandfine-grained(grammar)competenceand(sys-tem) performancemeasures—like the num-berof readingsobtainedpertestitem,varioustime and memory usagestatistics,ambigu-ity andnon-determinismmetrics,andsalientpropertiesof the resultstructures—andstorethemin auniform,platform-independent dataformatasacompetenceandperformancepro-file; and> graphical facilities to inspect the resultingprofiles, analyze system competence(i.e.grammatical coverage and overgeneration)andperformance(e.g.cpu time andmemoryusage,parsersearchspace,constraintsolver

Parser 3Parser 2

Parser 1

Grammar 3Grammar 2

Grammar 1

TestSet 3Test

Set 2TestSet 1

ParallelVirtual

Machine

C and Lisp API

RelationalDBMS

Batch Control

StatisticsUser

Interface

ANSI C Common-Lisp Tcl/Tk

Figure 1: Roughsketch of [incr tsdb()] architec-ture: thecoreenginecomprisesthedatabaseman-agement,batchcontrol and statisticscomponent,andtheuserinterface.

workload, and others)at variablegranulari-ties, aggregate, correlate,and visualize thedata, and compareamongprofiles obtainedfrom previousgrammaror systemversionsorotherprocessingenvironments.

As it is depictedin Figure 1, the [incr tsdb()]architecturecanbe broken down into threemajorparts:(i) theunderlyingdatabasemanagementsys-tem (DBMS), (ii) the batchcontrol and statisticskernel (providing a C and Lisp applicationpro-gram interface to client systemsthat can be dis-tributedacrossthe network), and(iii) the graphi-caluserinterface(GUI). Although,historically, theDBMS wasdevelopedindependentlyandtheker-nelcanbeoperatedwithout theGUI, thefull func-tionality of the integratedcompetenceandperfor-mancelaboratory—asdemonstratedbelow—onlyemerges from the combinationof all threecom-ponents.Likewise, the flexibility of a clearly de-fined API to client systemsandits ability to par-allelize batchprocessingand distribute test runsacrossthenetwork have greatlycontributedto thesuccessof thepackage.Thefollowing paragraphsreview someof the fundamentalaspectsin moredetail,sketchessentialfunctionality, andcommentonhow they havebeenexploitedin theDFKI –YYcooperation.

Abstraction over Processors The [incr tsdb()]environment, by virtue of its generalizedpro-file format, abstractsover specificprocessingen-vironments. While grammar engineersin theDFKI –YY collaborationregularly use both theLKB (primarily for interactive development)andPET (mostly for batchtestingand the assessment

17

Page 22: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

of results obtainedin the YY production envi-ronment),usageof the [incr tsdb()] profile anal-ysis routinesin most aspectshides the specificsof thetoken processorusedin obtaininga profile.Both platformsinterpretethe sametyped featurestructureformalism, load the sameset of gram-marsourcefiles, and(unlessmalfunctioning)pro-duceequivalentresults.Using [incr tsdb()], gram-marianscanobtainsummaryviews of grammati-cal coverageandovergeneration,inspectrelevantsubsetsof theavailabledata,breakdown analysisviews accordingto variousaggregation schemes,andzoom in on specificaggregatesor individualtest items as appropriate. Moreover, processingresultsobtainedfrom the (far moreefficient) PET

parser(thathasnovisualizationor debuggingsup-portbuilt in), oncerecordedasan[incr tsdb()] pro-file, canbeusedin conjunctionwith theLKB (con-tingenton theuseof identicalgrammars),therebyfacilitatinggraphicalinspectionof parsetreesandsemanticformulae.

Parallelization of TestRuns The[incr tsdb()] ar-chitecture(seeFigure1) separatesthe batchcon-trol andstatisticskernel from what is referredtoasclient processors(i.e. parsingsystemslike theLKB or PET) throughanapplicationprograminter-face(API) andtheParallelVirtual Machine(PVM;Geist, Bequelin, Dongarra, Manchek, & Sun-deram,1994)message-passingprotocollayer. Theuseof PVM—in connectionwith taskscheduling,error recovery, androll-over facilities in the [incrtsdb()] kernel—enablesdevelopersto transparentlyparallelizeand distribute executionof batchpro-cessing.At YY, grammarianshada clusterof net-workedLinux computeserversconfiguredasasin-gle PVM instance,sothatexecutionof a testrun—usingtheefficient PET run-timeengine—couldbecompletedasa matterof a few seconds.Thecom-binationof near-instantaneous profilecreationand[incr tsdb()] facilitiesfor quick,semi-automatedas-sessmentof relevant changes(seebelow) enableddevelopersto pursuea stronglyempiriciststyleofgrammarengineering,assessingchangesandtheireffects on actualsystembehavior in small incre-ments(oftenmany timesperhour).

Structur ed Comparison One of the facilitiesthat has proven particularly useful in the dis-tributed grammarengineeringsetup outlined inSection2 above is theflexible comparisonof com-petenceandperformanceprofiles.The[incr tsdb()]packageeasescomparisonof results on a per-

item basis, using an approachsimilar to Un ? xdiff(1), but generalizedfor structureddatasets.By selectionof a set of parametersfor intersec-tion (andoptionally a comparisonpredicate),theuser interfaceallows browsing the subsetof testitems (and associatedresults) that fail to matchin the selectedproperties. One dimensionthatgrammariansfound especiallyuseful in intersect-ing profilesis on thenumberof readingsassignedper item—detectingwhere coveragewas lost oradded—andon derivation trees(bracketedstruc-tureslabeledwith rulenamesandidentifiersof lex-ical items)associatedwith eachparseranalysis—assessingwhereanalyseshave changed.Addition-ally, using a user-suppliedequivalencepredicate,the sametechniquewas regularly usedat YY totrack theevolution of meaningrepresentations(asthey formtheinterfacefrom linguisticanalysisintothe back-endknowledgeprocessingengine),bothfor all readingsandtheanalysisrankedbestby theparseselectionheuristics.

Zooming and Interacti ve Debugging Inanalysing a new competenceand performanceprofile,grammarianstypically startfrom summaryviews (overall grammaticalcoverage,say), thensingle out relevant (or suspicious) subsetsofprofile data, and often end up zooming in tothe level of individual test items. For most [incrtsdb()] analysisviews the ‘success’criteriacanbevaried accordingto user decisions:in assessinggrammaticalcoverage,for example, the scoringfunction can refer to virtually arbitrary profileelements—rangingfrom the mostbasiccoveragemeasure(assigningat leastone reading)to morerefinedor application-specificmetrics,theproduc-tion of awell-formedmeaningrepresentation,say.Although the generalapproachallows outputan-notationsonthetestdata(full or partialconstituentstructuredescriptions,for example),developerssofar have found the incremental,semi-automatedcomparisonagainstearlierresultsamoreadequatemeansof regressiontesting. It would appearthat, especially in an application-driven andtightly scheduledengineeringsituation like theDFKI –YY partnership, the pace of evolutionand generallack of locality in changes(seetheexamplesdiscussedin Section 2) precludestheconstructionof a static, ‘gold-standard’target forcomparison. Instead,the structuredcomparisonfacilities of [incr tsdb()] enable developers toincrementallyapproximatetargetresultsand,even

18

Page 23: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

12-sep-2001 (13:24 h) – 14-feb-2002 (17:14 h)

404550556065707580859095

Grammatical Coverage (Per Cent)

(generated by [incr tsdb()] at 29-jun-2002 (20:49 h))

@ @ @ @ @@ @ @@ @ @ @ @@ @@ @ @@@@@ @@ @@@ @@ @ @ @ @ @@@@ @@ @ @@ @@

@ @@ @@@ @ @ @A@ @ @ @@B@ @ @@@ @@@ @ @ @ @@ @ @@B@ @ @

C C C CCC C C C C C C C CC C C C C C CC C C C CC C C CCC C CDCCCC CC C CCC CC C CCC CC

E — ‘banking’F — ‘trading’

12-sep-2001 (13:24 h) – 14-feb-2002 (17:14 h)

0102030405060708090

Ambiguity (Average Number of Analyses)

(generated by [incr tsdb()] at 29-jun-2002 (20:59 h))

@ @ @ @ @@ @ @ @ @G@@ @@ @ @ @ @@@ @ @ @@ @ @ @@B@ @ @@ @ @ @ @ @ @ @C

CC C C C C CC

C C C C C C CHCCCC C C C C CCC C CE — ‘banking’F — ‘trading’

Figure2: Evolutionof grammaticalcoverageandaverageambiguity(numberof readingspertestitem)overafive-monthperiod;‘banking’ and‘trading’ aretwo datasets(of some700and400sentences,respectively)of domaindata.

in a highly dynamicenvironmentwheregrammarand processingenvironment evolve in parallel,track changesand identify regressionwith greatconfidence.

4 Looking Back—Quantifying EvolutionOver time, the [incr tsdb()] profile storageaccu-mulatesprecisedataon thegrammardevelopmentprocess. Figure 2 summarizestwo aspectsofgrammaticalevolutioncompiledoverafive-monthperiod (and representingsome130 profiles thatgrammariansputasidefor futurereference):gram-maticalcoverageover two representative samplesof customerdata—onefor anon-linebankingap-plication, the otherfrom an electronicstocktrad-ing domain—iscontrastedwith the developmentof global ambiguity (i.e. the averagenumberofanalysesassingedto eachtest item). As shouldbe expected,grammaticalcoverageon both datasetsincreasessignificantly as grammardevelop-mentfocuseson thesedomains(‘banking’ for thefirst threemonths,‘trading’ from thereon). Whilethecollectionof availableprofiles,apparently, in-cludesa numberof datapoints correspondingto‘f ailed’ experiments(fairly dramaticlossesin cov-erage),thelargerpictureshows mostlymonotonicimprovement in coverage. As a control experi-ment, the coveragegraph includesanotherdatapoint for the‘banking’ datatowardstheendof thereportingperiod. Two monthsof focusseddevel-opmenton the ‘trading’ domain have not nega-tively affectedgrammaticalcoverageon the data

setusedearlier. Correspondingto the (desirable)increasein coverage,thegraphon theright of Fig-ure 2 depictsthe evolution of grammaticalambi-guity. As hand-built linguistic grammarsput greatemphasison the precisionof grammaticalanaly-sis andthe exclusionof ungrammaticalinput, theoverall averageof readingsassignedto eachsen-tencevariesaroundrelatively smallnumbers.Forthe moderatelycomplex email data2 the grammaroften assignslessthan ten analyses,rarely morethan a few dozens. However, not surprisinglythe additionof grammaticalcoveragecomeswitha sharpincreasein ambiguity (which may indi-cateovergeneration):thegraphsin Figure2 clearlyshow that,oncecoverageon the‘trading’ datawasabove eighty per cent, grammariansshifted theirengineeringfocuson‘tightening’ thegrammar, i.e.theeliminationof spuriousambiguityandovergen-eration(seeSiegel & Bender, 2002,for detailsonthegrammar).

Another view on grammar evolution is pre-sentedin Figure 3, depicting the ‘size’ of theJapanesegrammarover the samefive-monthde-velopmentcycle. Althoughmeasuringthe sizeof

2Quantifying input complexity for Japaneseis a non-trivial task,asthecountof thenumberof input wordswoulddependon theapproachto stringsegmentationusedin a spe-cific system(the fairly aggressive tokenizerof ChaSen,Asa-hara& Matsumoto,2000,in our case);to avoid potentialforconfusion,we reportinputcomplexity in the(overtly system-specific)numberof lexical itemsstipulatedby the grammarinstead:around50 and80, on average,for the ‘banking’ and‘trading’ datasets,respectively (asof February2002).

19

Page 24: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

12-sep-2001 (13:24 h) – 14-feb-2002 (17:14 h)

8800

9000

9200

9400

9600

9800

10000

10200

889092949698

100102104106

Grammar Size

(generated by [incr tsdb()] at 30-jun-2002 (16:09 h))I I II I I I I I I II I II III I I I II II I I IJIIIII II I I I I I I IIIII I I IKI I I II II III I II I I I II I II I I

II I I II I I I I I II I III I I I II I I I I II I II II I ILIIII II I III II II IIIIM M M MMM MM M M MNMM MMM M M MMMM MJM M MM MMM M MOMMM MM MMM M M M M M MM MMM M MM M M

M MM M M"MMM M M M M M M MMM MM M MMMNM MPMM M MQM MRM MM M MMM M M MJMM M M M M MM M MM MM M MLMMMM MM M MMM MM MM MMMMS — typesT — rules

Figure3: Evolution of grammarsize(in thenum-bers of types, plotted againstthe left axis, andgrammarrules,plottedagainsttheright axis)overafive-monthperiod.

computationalgrammarsis a difficult challenge,for theHPSG framework two metricssuggestthem-selves: the numberof types (i.e. the size of thegrammaticalontology) and the numberof gram-marrules(i.e. theinventoryof constructiontypes).As would be expected, both numbersincreasemoreor lessmonotonicallyover the reportingpe-riod, wherethe shift of focus from the ‘banking’into the ‘trading’ domainis marked with a sharpincreasein (primarily lexical) types. Contrastedto the significant gains in grammaticalcoverage(a relative improvementof morethanseventy percenton the ‘banking’ data),the increasein gram-mar size is moderate,though: aroundfifteen andtwenty per centin the numberof typesandrules,respectively.

5 ConclusionsAt YY andcooperatingpartners(primarily DFKISaarbrucken and CSLI Stanford), grammarians(for all languages)aswell asdevelopersof boththegrammardevelopmenttoolsandof theproductionsystemall usedthe competenceandperformanceprofiling environmentas part of their daily engi-neeringtoolbox. The combinationof [incr tsdb()]facilities to parallelizetest run processingand abreak-throughin client systemefficiency (usingthePET parser;Callmeier, 2002)hascreatedanex-perimentaldevelopmentenvironmentwheregram-marianscanobtainnear-instantaneous feedbackontheeffectsof changesthey explore.

For the Japanesegrammar specifically, thegrammardevelopersat both endswould typically

spendthefirst tento twentyminutesof thedayob-taining fresh profiles for a numberof sharedtestsetsanddiagnosticcorpora,therebyassessingthemostrecentsetof changesthroughempiricalanal-ysisof their effects. In conjunctionwith a certainrigor in documentationandcommunication,it wasthe ability of both partnersto regularly, quickly,and semi-automaticallymonitor the evolution ofthe joint resourcewith greatconfidencethat hasenabledtruly parallel developmentof a single,sharedHPSG grammaracrosscontinents. Withina relatively short time, the partnerssucceededin adaptingan existing grammarto a new genre(email rather than spoken language)and domain(customerservicerequestsratherthanappointmentscheduling),greatly extending grammaticalcov-erage(from initially aroundforty to above ninetyper centon representative customercorpora),andincorporatingthe grammar-basedanalysisengineinto a commercialproduct. And even thoughinFebruary2002,for businessreasons,YY decidedto reorganizegrammardevelopmentfor Japanese,the distributed,parallelgrammardevelopmentef-fort positively demonstratesthat methodologicaland technologicaladvances in constraint-basedgrammar engineeringhave enabledcommercialdevelopmentand deployment of broad-coverageHPSG implementations,a paradigmthat until re-centlywasoftenbelievedto still lack thematurityfor real-world applications.

Acknowledgements

TheDFKI –YY partnershipinvolvedalargegroupof peopleat both sites. We would like to thankKirk Oatman,co-founderof YY and first CEO,andHansUszkoreit, ScientificDirector at DFKI,for their initiative and whole-heartedsupport tothe project; it takesvision for both corporateandacademictypesto jointly develop an open-sourceresource. Atsuko Shimada(from SaarlandUni-versity), aspartof a two-monthinternshipat YY,hasgreatlycontributedto thepreparationof repre-sentative datasamples,developmentof robustpre-processingrules,andextensionsto lexical cover-age.Our colleagueandfriend Asahara-san(of theNara AdvancedInstitute of Technology, Japan),co-developerof theopen-sourceChaSentokenizerandmorphologicalanalyzerfor Japanese,wasin-strumentalin the integration of ChaSeninto theYY productandalsohelpeda lot in adaptingand(sometimes)fixing tokenizationandmorphology.

20

Page 25: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ReferencesAsahara,M., & Matsumoto,Y. (2000). Extended

modelsandtoolsfor high-performancepart-of-speechtagger. In Proceedingsof the 18th In-ternationalConferenceonComputationalLin-guistics(pp.21–27). Saarbrucken,Germany.

Butt, M., King, T. H., Nino, M.-E., & Segond,F.(1999). A grammarwriter’s cookbook.Stan-ford, CA: CSLI Publications.

Callmeier, U. (2002).Preprocessingandencodingtechniquesin PET. In S.Oepen,D. Flickinger,J. Tsujii, & H. Uszkoreit (Eds.), Collabora-tive language engineering. A casestudyin ef-ficient grammar-basedprocessing. Stanford,CA: CSLI Publications.(forthcoming)

Carroll, J. (1994). Relatingcomplexity to practi-calperformancein parsingwith wide-coverageunification grammars. In Proceedingsof the32ndMeetingof theAssociationfor Computa-tional Linguistics(pp.287–294). LasCruces,NM.

Copestake, A. (2002). Implementingtypedfea-ture structure grammars. Stanford,CA: CSLIPublications.

Erbach,G. (1991). An environment for exper-imenting with parsingstrategies. In J. My-lopoulos & R. Reiter (Eds.), ProceedingsofIJCAI 1991(pp. 931–937). SanMateo,CA:MorganKaufmannPublishers.

Erbach,G., & Uszkoreit, H. (1990). Grammarengineering. Problemsandprospects(CLAUSReport# 1). Saarbrucken,Germany: Compu-tationalLinguistics,SaarlandUniversity.

Flickinger, D., Copestake,A., & Sag,I. A. (2000).HPSG analysisof English. In W. Wahlster(Ed.), Verbmobil. Foundationsof speech-to-speech translation(Artificial Intelligenceed.,pp.321–330). Berlin, Germany: Springer.

Geist,A., Bequelin,A., Dongarra,J.,Manchek,W.J.R., & Sunderam,V. (Eds.). (1994). PVM —parallel virtual machine. Ausers’ guideandtu-torial for networkedparallel computing. Cam-bridge,MA: TheMIT Press.

Makino, T., Yoshida,M., Torisawa, K., & Tsu-jii, J. (1998). LiLFeS — towardsa practicalHPSGparser. In Proceedingsof the 17th In-ternationalConferenceonComputationalLin-guistics and the 36th Annual Meetingof the

Associationfor ComputationalLinguistics(pp.807–11). Montreal,Canada.

Muller, S., & Kasper, W. (2000). HPSGanaly-sisof German.In W. Wahlster(Ed.),Verbmo-bil. Foundationsof speech-to-speech transla-tion (Artificial Intelligenceed.,pp.238–253).Berlin, Germany: Springer.

Oepen,S., & Callmeier, U. (2000). Measureformeasure: Parsercross-fertilization.Towardsincreasedcomponentcomparability and ex-change.In Proceedingsof the6thInternationalWorkshopon Parsing Technologies (pp. 183–194). Trento,Italy.

Oepen,S., & Flickinger, D. P. (1998). Towardssystematicgrammarprofiling. Testsuitetech-nology ten yearsafter. Journal of ComputerSpeech andLanguage, 12(4) (SpecialIssueonEvaluation), 411–436.

Oepen,S.,Netter, K., & Klein, J. (1997). TSNLP

— TestSuitesfor NaturalLanguageProcess-ing. In J.Nerbonne(Ed.),LinguisticDatabases(pp. 13–36). Stanford,CA: CSLI Publica-tions.

Penn,G. (2000). Applying constrainthandlingrulesto HPSG. In Proceedingsof thefirst in-ternationalconferenceon computationallogic(pp.51–68). London,UK.

Siegel,M. (2000).HPSGanalysisof Japanese.InW. Wahlster(Ed.), Verbmobil.Foundationsofspeech-to-speech translation(Artificial Intelli-genceed., pp. 265–280). Berlin, Germany:Springer.

Siegel, M., & Bender, E. M. (2002). Efficientdeepprocessingof japanese.In Proceedingsof the19th InternationalConferenceon Com-putationalLinguistics.Taipei,Taiwan.

Uszkoreit, H., Backofen, R., Busemann,S., Di-agne,A. K., Hinkelman, E. A., Kasper, W.,Kiefer, B., Krieger, H.-U., Netter, K., Neu-mann, G., Oepen, S., & Spackman,S. P.(1994). DISCO — an HPSG-basedNLPsystem and its application for appointmentscheduling.In Proceedingsof the15thInterna-tional Conferenceon ComputationalLinguis-tics. Kyoto,Japan.

21

Page 26: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

UWVYX[Z]\[^K_a`cbed[fcgihkjclnmNo+prqcset[ucvxw�y]z|{|})~+�n�-����� �a�K�������K�����P�J�K���r���r�����)�+�¡ £¢�¤¦¥¨§)©«ª�¬+­¯®°²±O³µ´A¶¸·�¹ º+»G¼¾½[¿ÀÂÁÄÃÆÅKǾÈKÉ¡ÊÌËÎÍÐÏ¡Ñ£ÒAÓÆÔNÕÆÖ×cØÚÙ�ÛÂÜÞݵßKàâáNã�äæåÎçéèµêëÐìAí�îðï¡ñÎò�ó¦ôöõ¾÷RøÚù«úxûiü¾ýþÚÿ������������� ������������������ �"!

#%$'&(*),+.-0/

13254"6�798�:�;=<0>@?BA,C5D"EGF"HGIKJ�L0M5N OQP.R S,T�U=V�WYX,Z5[ \Q]^�_,`�aKb�c�d,e=fhg�i�jBk=l�mon,prqtsBu�vxw�yz"{G|=}Y~0���*���=���Y�Q�����*��=� ���t�B�B�=�0� ����������t���o �¡=¢ £o¤�¥�¦¨§�©�ª¬« ­t®Y¯°@±0²B³=´Bµ�¶=· ¸º¹�»0¼o½¿¾ À�Á"Â*ÃÅÄ�Æ,Ç È É,Ê�Ë=Ì�Í�Î,Ï5Ð Ñ"ÒÓGÔ�Õ×ÖtØrÙ9ÚrÛÝÜ"Þàß=á0âäãKåQæQçºèêéBë¿ì�íïîoð¬ñtòrótôöõh÷�ø=ùBú=û�üýBþ=ÿ������������� ����������������������� �!#"%$�&�'%(*),+�-/.102�3*4�56798;:=<�>@?ACBEDGF�H�IKJ*LNMPO*Q�R9S�TKU�V�W9X1Y[Z]\�^`_ba�c/dfeg�h%ikjl�m*npo�q9rPsGtKu�v�w[x;yEz�{}|%~1�@�[�p����/�;�*�=���=�/�1�����P������������K�*���E�K���P� ¢¡9£�¤¦¥E§¨�©«ªE¬C­�®�¯�°²±´³�µ·¶¹¸fº¼»K½¾=¿kÀ�ÁCÂ�Ã%Ä�ÅPÆ Ç[È9É Ê�ËkÌ9Í1Î%ϼÐ�Ñ�ÒPÓ Ô�Õ×ÖCØ´Ù,Ú%Û@ÜÝ=Þ�ßpà/á@â�ã=ä�åfæèç�éëê�ìîíï�ðñ�ò�óõôPöN÷Cø@ùEúüû*ýÿþ������������� �������������������! #"%$'&)(�*,+�-�.0/'132�4�5�6�798:<;>=)?�@BADC,E'FG�HJILK�M�N#O<P!QSRBTVU#W,XZY\[^]�_'`a3bdcfe)g�h�iBj�k�l�m#nporqLs!tvuBwxBy'z|{,}~#�#�������������J�'�)���,�S�,�V�,�L�<���<���!�����������J��D ¡B¢�£�¤�¥9¦�§¨D©«ª�¬�­J®p¯<°'±�²³�´�µ�¶·<¸J¹#º'»½¼�¾�¿�À�Á!ÂVÃÅÄ#ÆÇfÈ�É�Ê�Ë)Ì�ÍÎ'Ï«Ð^Ñ�ÒJÓJÔÖÕ�×ÖØ,Ù#Ú'ÛJÜ�ÝJÞ�ß�à'áJâ#ãZä�åræ!ç�èLé�ê<ëì^íîðï�ñfòpóôrõ!ö÷�ø�ù�ú<û%üfýJþ«ÿ������������ �� ����������� ���

� �����! "$#�%'& (!)+*-,.0/ 132�4 57698;:=<?> @BA�C�DFEGIHKJ0L�MFN�OQPSRUT9VWFX;Y=Z�[9\^]`_Kacb9d�egf+hikj�l�m�n!ocp�q r�sut`vxw�y{z}|S~����U�;���������;�c�;�Q�;���Q�����7���Q�g������7������ �¡£¢�¤ ¥§¦g¨F©{ª7«c¬7­�®`¯g°²±�³�´�µS¶Q·¹¸»º�¼U½»¾À¿ Á�Â}Ã}Ä{Å}ÆÇ`È7É Ê�Ë�ÌFÍgÎgÏKÐ'ÑÓÒ ÔÖÕ�× Ø;Ù�ÚSÛÜÞÝ�ß�à�á`âQãåä+æèç�éQê+ë7ì+ícîðïòñ`óõôðö÷ùø0úüûòý}þ�ÿ ��������� � ����������� ����� ����� �"!$#&%'�(*)�+-,/.10�243-5�687-9;:�<>= ?�@�A4BDC�E8F�G�H I�JLK>M;N�OQPSRUT>V�WYX[Z�\�]�^&_a`bdc�e$f�g�h�i;j�k�l*m&nporq s&tvuSw�xSy zL{ |D} ~��������������/�Q�Y������������S�������������8 ¢¡S£�¤>¥�¦8§�¨;©�ªD«�¬1­S®�¯�°�±;²[³L´¶µ�·S¸-¹»ºY¼�½"¾�¿1À�ÁÃÂ�ÄÅSÆÈÇ>É�Ê&ËÍÌ>Î�ÏÑÐ�ÒSÓ�Ô�Õ;ÖY×�Ø>Ù�ÚÜÛDÝ�Þ�ßÜàâáäã�å&æ¶ç�è é ê/ëÜìâí�î�ïð�ñ�ò�óÃô>õSö�÷�ø;ù�úLûSü/ýÿþ������������� ������������������ ��!#"%$'&)(�*,+.-�/102�3#465 7�8:9<;�=:>@?�ACB�DFEHGCIKJ#LNMPORQTS�U�V�WKX�Y'Z�[�\^]#_�`�acbed�fhghi%jlkm#n�oqpPrtsvuCw�x�yRzq{�|~}��������~�����#�:�������������:���C�P�%�)�~�����C�������v ¢¡�£�¤¥P¦¨§#©vª¢«­¬¯®�°@±�²K³h´6µq¶�·­¸)¹�º�»½¼)¾À¿�ÁlÂÄÃCÅÄÆ ÇCÈ�ÉlÊ�Ë Ì�ÍÏÎKÐÒÑÓ�Ô#Õ%Ö)×KØ#ÙCÚKÛhÜ'Ý�Þlß:à�á�â�ã�äæåKç�è é�êìë�í î�ï�ðvñ#òôó õCö�÷�øcù1ú�ûýü�þ#ÿ������������ ���������������� �"!$#&%('*),+�-/.10�2 3�4�576�8:9 ;&<>=@?A�BDC�E:F,G�HIDJ�K/L/MON�PRQ�S�TOU�V$WYX[Z]\O^�_1`/acb$d�e�fhg�i&j�k�lnm�oprqts�uOv�w�x y{zO|~}�:�����:���,�"�����$���������1��� ������� �O�������¡ �¢¤£�¥ ¦¨§�©Oª¬«$­¯®:°�±�²�³�´/µO¶·t¸$¹Oº�»,¼ ½¿¾OÀnÁÃÂÅÄÇÆ�È,ɯÊ�Ë¿Ì,ÍÏÎÑÐ�Ò�Ó/Ô$Õ¨Ö*×OØ¿Ù�Ú/ۨܬÝ:Þ&ß à/á�âOãÑäåçæOè�é�êOë�ìîí�ï$ðòñôó�õ�ö�÷�ø�ù�ú�û�üRýOþ/ÿ���������� �������������������� !�"$#&%$'�()+*�,$-/.10�24365�798�:<;&=?>A@CB�DFEHG�IKJ�LNM�OQP�RTSVU�W9XZY�[�\$]_^`Aa�bVcedgfihKjlk�monep�q�r_sltiuwv1xyz�{<|1}�~g���e�����A�������V�g�&�i���l���l�e�g�

�1���A�l�i�9�$� ��¡H¢H£ ¤�¥§¦©¨�ª�«&¬9­©®9¯l°4±A²N³1´+µ�¶¸·e¹/º�»½¼A¾&¿9ÀÂÁ�Ã�ħÅ&Æ�ÇgÈÉ1Ê�ËAÌlÍiÎ9Ï$РÑQÒAÓ�Ô�Õ×ÖiØ�Ù?Ú ÛwÜ�Ý�Þ ßáà?âNã�ä åVæèç&éëêiì+íïî+ðeñlò<óõô�ö_÷_ø�ùú1û�ü×ý$þ�ÿ���������� ����������� ��������� �"!$#%�&(' )+*-,�.0/2143�576(8:9<;=?>A@(B(C(DFE

G?HJILK0M�NPO�Q�R S�TUWV4XZY([+\^]�_<`ba cedLf gih�jAk(l"mon�p0qrts�uLvxw(y�z|{ }�~��L�x������� ��^�?�L���0���0���<���7���A���A������� �t �¡(¢¤£A¥¦�§x¨ª©L«�¬ ­"®°¯�±A²´³xµA¶0·�¸(¹"º°»+¼¾½:¿(À$Á<Â�Ã<Ä�Å�ÆÈÇ+ÉËÊ0Ì�Í�ÎÈÏ�Ð+Ñ�Ò<Ó"Ô<Õ�Ö×AØoÙLÚ(Û<Ü�ÝßÞáà<âã�äæå-ç(è+é ê�ë:ì�íïî�ð�ñóò�ô0õ÷ö�ø:ù0ú�û(ü�ý�þ�ÿ������������ ������������������! #"�$&%('()+*-,/.1032+4-57698;:�<>=�?�@BA!CED�F+G/HEIEJK�L�M N;O�P(Q R�S TVU�WEX�Y�Z([ \!]�^ _!`ba�c&d�e�f(g;h i!j�kl7m+npo�q�r�s�t�u+v�w;x�y�z!{�|!}�~��(�p���+���������������������(������������������������� ¢¡�£�¤!¥§¦�¨!©9ª(«�¬-­/®°¯�±�²´³�µ9¶;·(¸(¹�º(»�¼!½�¾¿9ÀÂÁ!Ã�Ä!Å�Æ�Ç(ÈpÉËÊbÌ�ÍÎ!Ï9Ð�ÑbÒ�Ó+ÔÖÕ�×bØ´Ù9ÚÜÛ�Ý-Þ�ß�àpá9â-ã�äVå�æ�çEè�é�êìë�íïîñð�ò+óõô;ö&÷ùøûú+ü-ýÿþ������������ ���������� �������! "�#$�%�&�'�(�) *,+�- .0/2143�56�798;:=<�>2?!@�A;B�C!DFE2G0HJI�K�L�MONQP4R�SUTQV�WOXZY\[^]_a`0bdcfeOgihkjOl�m=n4oFp\qFr�s\tu,v�w\xiy{z!|�}U~f�{�����4���d���������,���J�������\���d�����,�����J�� ,¡�¢�£�¤¥O¦=§�¨Q©�ª¬«­;®¯�°�±2²�³J´�µ�¶�·¸�¹0º¼»f½4¾k¿�À�ÁUÂ�Ã�Ä�Å�Æ�ÇÈ�É0Ê�Ë�Ì�Í�Î4ÏfÐÑ�Ò�Ó�ÔdÕ×Ö¼Ø9Ù�Ú=ÛfÜFÝUÞ�ß,àâáUã,ä,å�æ,ç�èêé^ëdìJí2î9ï

ð�ñ=òôódõ÷ö�ø4ù\ú�ûUüký�þ�ÿ���������� ����������������������! �"$#&%('*)�+-,/.02143&57698;:�<�=?>(@�A�BDCFE7G!H�I�JLK MON!PRQ�SUTLV�W9X(Y[Z�\]&^�_O`Da;bdc9e*fg�h&i�jlk�m!n�o�p�q�r s-t�u�v!wyx�z�{�|(}D~��������!���(���&�?��/�����d�����*������7��������/����� ��¡�¢�£�¤�¥/¦¨§*©«ªL¬�­¯®2°4±&²7³9´;µ·¶$¸º¹¼»L½�¾�¿�À�ÁºÂ(Ã�ÄÅ/Æ�Ç/È�É�Ê�ËÍÌ�ÎÐÏ2Ñ�Ò¼Ó�ÔÖÕ�×2Ø[Ù9Ú�Û�Ü�ÝßÞáàãâ�ä&å�æ�ç(è�édê�ëãìLí�î-ï�ðòñ�óô*õ!ö�÷$ø�ùLú�û ü!ýÖþ¼ÿ���������� ���������������������� "!$#&%('�)�*,+.-0/213 465.7�8:9<;�="><?A@AB�CED$FHG�IAJ�KMLHN�O�P�QSRHT�UAV�W�X�Y[ZA\�]"^�_a`�b�ced&f�gih

j�k�lnmpoiqir:sut�v�wyx{z�|e}{~A�����E�$�:���0�����e�����.�e���n���.�(�"���E���A�$��H����� �¢¡ £�¤¦¥�§i¨�©eª�«�¬E­�®p¯y°�±³²µ´2¶¸·�¹»º&¼�½�¾i¿¸ÀÂÁ Ã�ÄÆÅ&ÇÉÈ,ÊeËÌ�Í�Î�ÏÑÐyÒ.Ó$ÔAÕ�Ö�×eØAÙ�Ú�Û¢Ü:Ý�Þ�ß à�á�âiã�ä$å(æ

3 ç6è.é�ê:ëíìHî"ïíðEñÉòeóõôõöy÷õø�ù�ú�ûAü�ý�þ�ÿ���������� ���������������������� �"!$#�%'&'(*) +-,/.103254768:9;=<'>$?�@BA�C�D*E�FHGJI�KMLON'P Q=RTSVU=W

XZY\["]�^`_ba"cVd=e/fhgbi/jlk�m`n�ohp qsr"tOuwv/xzy/{�|/}�~����M�����s�"���"���7���b��b�"�V�=�/�h�b�M�����Z�����b  ¡h¢�£"¤M¥"¦b§�¨ª©�«B¬7­�®°¯*±z²s³ ´/µ7¶¸·º¹"»z¼½º¾�¿�ÀbÁÃÂ"ÄÆÅ7ÇMÈ�ÉhÊÌËbÍhÎ�ϸÐ�ÑzÒzÓ�ÔbÕMÖ ×ÙØ�ÚÃÛ�ÜÞÝzß"à�áãâ"äæå�ç�è�é�ê1ëbì"íïî�ð/ñ"òOóMô õöZ÷�ø�ùbúïû üsý"þ"ÿ���������� ������������������� �"! #�$�% &('*),+�-�.0/ 1 2354060708:9<;=?>�@BA�CED�F�GIH�JLK*M�NPORQ S�TVU�W<XLY Z\[�]�^`_ba cLd�e�fPg�h<i�jlk0m�npo(q�r�s0t�u0v0wx y�z0{<|�}<~�� ���������0���������������"���L��� � �����������0���  ¡p¢5£ ¤:¥§¦�¨ª©*«�¬I­®b¯°`±:²�³�´¶µb·�¸<¹0º�»½¼�¾½¿�ÀÂÁ�à Ä*Å�Æ<Ç È�É�ÊÌË*ÍÂÎ�Ï�Ð\Ñ�Ò ÓÔÖÕ<×�Ø�Ù�Ú Û�Ü�ÝÞ ß�à0á<â�ã<ä�å æ�ç�è�é�ê0ëìIí�î0ï�ð�ñ:òLó�ôõ÷ö�ø ù�ú û�ü�ý�þ�ÿ���������� ���������������������! �"#%$'&)(+*-,'.0/2143658769�:�;�<>=4?�@0A�B�CED4F�GIH�J>K4LNMPO-Q%RPS0TVUXWZY\[^]4_�`\aEbc6d�e-f�g�h\i4j4k%l�m�n�o�p4qsrtvu�w�xzy�{I|�}z~I�4���8�+�������-���E���������6�0�����I���������6�����6���� >¡�¢'£z¤E¥�¦§�¨�©�ª�«�¬'­¯®z°\±�²´³�µ·¶0¸V¹'ºI»\¼4½V¾À¿0ÁVÂ�ÃEÄ�ÅVÆ6Ç�È�É�ÊzË\Ì�ÍÏίÐ�Ñ4ÒVÓ'Ô�Õ

Ö´×

Page 27: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Ø�Ù�ÚÛ>Ü>ÝVÞ�ßáà�âäãæå%çäè%é�ê�ëíìIîäïíð%ñò�óVôäõ�öø÷úù¯ûýüÿþ������������ ��������������������� "!�#%$'&)(+*�,�-/.1032+4�536�738�9�:<;>=@?�A�B�C�DFEGIH/JFK�LNM�OQP�RTS

U VXW<Y�Z\[^]�_�`\acb�dTe�f�g�hji�k�l�monFp�q�r�sTt�u�vxwFy�zQ{}|I~/�+���N������T�

�����������'���<�'���@�\���������+���� �¡)¢o£�¤¦¥�§�¨�©\ª�«�¬�­<®°¯�±�²o³�´�µ�¶F·¸�¹)º�»�¼N½�¾I¿ÁÀ<ÂÄÃ�Å�Æ�ÇÉÈIÊ@Ë�ÌFÍÏÎÑÐIÒ/Ó+Ô�ÕNÖ�×ÙØ�Ú�Û�Ü<Ý<ÞÁßáà\â�ãåäQæ@ç�èé+ê�ëåì\í�î<ï�ð�ñ\ò�ó�ô�õ�ö\÷)ø�ù�úNû�üIýÿþ���������� ������� �������������

� �! #"%$'&)(+*�,.-0/01�2%354�687:9';.<>=�?'@BADCE0F+G0H�IJ�KML�NO0PRQ�S8T�UWVYX[ZM\^]`_ba�ced�fhg ij�klnmpobqer0sutwvxwyWzp{�|}W~n���[���W�>�� �8� �����[�>�����b��� �%����n�������� W¡£¢¥¤�¦ §M¨�©eª«h¬b­�®�¯u°±³²[´µ·¶¸¹

º�»�¼'½>¾w¿WÀ Á�Â�ÃeÄÆÅÈÇ�É�Ê�ËWÌ>ÍÏÎ0Ð`Ñ�ÒÔÓ�Õ Ö>×pØ8Ù�ÚWÛ�ÜÞÝ�ßwà á�âeã�ähåbæWç8è.é�ênëìbíbîÔï�ð�ñ�ò'ó�ô�õ'ö£÷�ø�ùú�û`üWý�þ�ÿ������������� ��������������������! "$#&%�')(�*+-,/.10$2�354�61798;:�<>=@?�ACBED�FHGJI/K$L@M@N�O1P>QSR�TVU1WJX�Y�Z�[$\@]S^>_�`ba�ced�fhg>i�jlkm>npoJq r�s thuwv>x y�z {>|�}@~$�p�������5�$�@� ����� ���$�����w����p�9���������9���5�����@�� �¡£¢¥¤/¦¨§�©@ª«1¬>­¯®)°1±$²´³�µ�¶�·)¸º¹¼»&½e¾$¿ÁÀ�ÂÃ�Ä)Å�Æ1Ç@ÈwÉ-ÊJË>Ì@Í$Î�Ï�ÐeÑÓÒ>Ô1ÕHÖe×�Ø�ÙHÚ�ÛÝÜ�Þ1ßáà�âpã$äæå�ç�è�é9ê�ëpì!íJî�ï!ð�ñCò/óô�õ�ö/÷�ø1ùûú�üþýJÿ���������� ����������������������� �"!$#%�&�')(*,+.-0/�1325476,89;:$<.=�>@?.A5BDC.E,FHGIKJ�L�M�N�O�PQ"RTSVU�WYX$Z"[H\]^`_badc,e$f�g�h�ibjlkdmVnporqbsVtvu`w0x5y�zr{}|3~��`�D���������������b�V������������b�V�Y���)�V���d ¢¡£�¤�¥T¦�§�¨H©lªV«­¬¯®�°�±0²�³�´$µ0¶�·¹¸�º.».¼d½H¾0¿�À¯Á�ÂHÃHÄ$Å$Æ�ÇÈ,É.Ê0Ë�ÌÎÍ7ÏbÐÒÑ@ÓÕÔbÖV×�Ø$Ù0ÚVÛpÜ0ÝßÞ�à.áÒârã5äbåçæ�è�é�ê�ë�ì5íîTïVð�ñóò�ôõHö$÷ø5ù.úüûüý$þ�ÿ����������

�� ������������������! #"%$'&)(�*,+

-�.�/�0�1�2�3�46587�9�:<;�=>@?#ACB�D!E�FHG)I!J'KML'NPO�QMR�SUT�V�W#X@Y[Z[\,]_^ `�a<b�c�d#egf�h�i�jlkCm#n�o!p�qsrut�v@wyxUz){,|�}

~��

Page 28: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

���C�,�#�_�����!���������#��������M���������6�� �¡�¢s£¥¤§¦©¨�ª�«�¬�­s®�¯±°�²�³µ´,¶,·,¸y¹_º¼»©½�¾À¿)ÁCÂ�ÃyÄ�Å6ÆÇ�È6É�Ê#Ë)Ì©Í�Î�Ï6ÐÒÑ©Ó6ÔÖÕl×�Ø_Ù�ÚÜÛ�Ý�Þ�ßÖà�áãâ!äå!æ�ç�èCéëê,ì©í<î�ï!ð�ñ�òÒó�ô�õö�÷�ø�ù�ú#û�ü�ýPþ_ÿ���������� ��������� ����������� "!�#�$�%'&)(�*,+.-0/"132'4�5687�9;:=<?>A@"B�C"D3E"F?GIH�J8KML3N"O�P?Q�RTS8UWVYX Z\[^]�_�`"aYb0c�d.eYfhgji�klnm�o�p8qWrIs t�u0vxwAy?zj{.|"}�~ ���8�����8�W�����.�����3�?�����'�����0�W������� �¡W¢�£3¤�¥§¦¨I©«ª�¬�­¯®A°"±�²n³�´�µ\¶�·.¸W¹�º�»�¼"½�¾À¿8Á�ÂÄÃÆÅÈÇ�É�Ê�ËÍÌ�ÎÐÏ.Ñ�ÒÓ�Ô�Õ�Ö�×?ØÚÙ?ÛÝÜ0Þ�ß�àá�âWã ä å�æ?ç"è�éWê�ënì0íWî3ï8ð"ñWò�óõô öõ÷Wø�ùûúÐü�ýÿþ������������ ������������������� "!$#&%('�)�*,+.-�/�01�24365�798�:(;�<&=">@?BADC�EFHGI�JLK�MDN�OQP"RSUT�VXWBY[Z�\H]_^D`Da�b"cHd9e6fgih�j4k@lnmporq9sut$v&wyx�z�{p|~}H�������r�(���$�y�������y���H�U�p�D�������4�p���L����H� ��¡£¢,¤D¥D¦�§4¨p©«ª�¬�­�®°¯²±�³µ´�¶n·²¸�¹�ºD»�¼~½¿¾�À�ÁDÂ�Ã�ÄDÅ�ÆUÇ�È°ÉHÊ�ËÌ�Í(Î(Ï�Ð$ÑyÒÔÓBÕ�Ö6× ØLÙ6Ú@Û�ÜHÝ�Þàß�á�âuã¿ä6å£æ�ç[è¿é�êBë�ì�í�îðïHñXò[ó6ôöõ�÷pøúùpû&üþýÿ������������� �������������������� "!�#%$'&�(�)�*�+�,.-0/'1�24365"7�8�9;:=<>[email protected]�BDC�E.FHG4IKJHLDMNPOQSRUTWVYX.ZH[]\.^._a`Yb�c�dfePgihPjlk�monpYq4rHs�tvuxw�y�zi{�|~}D�4�.����������i���4�l�����i�H�������x�'���4���D��H���4 .¡�¢D£�¤.¥H¦4§©¨�ª�«K¬D­;®=¯±°%²H³�´xµ·¶x¸º¹¼»�½�¾.¿KÀ�ÁÃÂ�Ä�ÅÇÆ�ÈDÉ�Ê.Ë�̺Í�ÎÏÑÐ4ÒHÓÕÔiÖ�×Ø�Ù�Ú;Û%Ü�Ý�Þxß6àâáäã.å�æÑç�è�é%ê�ë�ìîíðïÑñHò�óõô�ö�÷xø�ù;ú=û�üþý�ÿ����������� ����� ����� ������������ "!$#&%('*),+.-0/$1�24365�7�8,9;:�< =�>@?BADC�E0F(GIH�JLKM0NIO;P;Q6RTS(U�V�W,XZY�[ \�]_^a`�b0c6d�egfIh�ikjml�npo*q�r;s t,ugv_w�xIy{z{|,}0~�����������&�4�0���$���m�;�Z���k�0�(�(�,�0���;���������;�(�_ �¡£¢m¤�¥4¦�§�¨�©�ª¬«�­&®�¯B°�±g²;³�´µ�¶�·¹¸�ºD»�¼�½Z¾I¿�À6Á�Â;ÃÅÄÇÆ È�É�Ê{Ë�Ì£Í,ÎÐÏ&Ñ£Ò�Ó$Ô$Õ�Ö ×IØ�Ù$Ú4ÛIÜ�Ý�ÞZßZà�á�â;ãä{å�æ¬ç&è�é�êìë�í�î6ï*ð,ñóò0ô�õ�öÇ÷Iø�ù�ú0ûýüZþ�ÿ���������� ����������������������! #"$�%'& (*)*+',�-/. 0�132 46537�8 9;:!<>=�?�@�A B�CED�F�GIH�J K3LM�N�O�P3QR�SUTWVX�Y�Z�[\3]_^�`>acbed�f�g�h�i�j�k3l_m3n�o!p!qsrutwv�xzy�{�|}�~������'������ ��������������������������������*������� �¡�¢3£W¤!¥�¦�§�¨�©�ª�«3¬­�®°¯6±>²³�´;µ�¶3·�¸ ¹º¼»½¿¾�ÀÂÁÄÃ>ÅÇÆÈÄÉ�Ê/ËeÌ*ÍÏÎ�ÐÒÑ�Ó�ÔÖÕ*×�ØÚÙ>ÛÝÜßÞ�à�á�âäãIå æèç�éwê�ë>ìÏí�îðï�ñ'òÏó>ôõ�ö�÷EøEù�ú û3üþý�ÿ���������� ��������������������� "!$#�%'&)(*,+"-�. /�0214365�798;:�<=?>A@�BDCFE�G�H�I2JLK$M�NPO�Q?RTS;U�V�WYX)Z�[F\ ]6^�_L` a�b�ced�f$g�h$i�j�k6l4mFn�op$qsrFtvu w�xLy{zv|�}�~��P� �F�����{�6�����������;���������s�6���;�;��� �������6 "¡�¢£�¤¦¥�§�¨�©9ª�«¬­6®°¯F±³²;´$µL¶¸·¦¹»º;¼½2¾ ¿ÁÀ$ÂÄÃ�Å$ÆÈÇ�É�Ê�Ë�Ì�Í ÎÏ?Ð$Ñ$Ò�Ó�Ô�Õ6ÖØ×ÙÛÚ�Ü�ÝßÞáàFâÛãLäæå�ç�èéëê¦ì»íïî�ð�ñYòTó�ô�õ÷ö»øØù�úüû9ýLþ�ÿ���������� ��������������������� �!#"�$&%('�)�*+-,�.0/01325476&8�9;:=<�>�?A@0BDCAE�FHG�IKJML-N0OQP�RHSUT�V�WYXKZA[H\�]�^`_5a�b�cDdfeg0h=i=j�k�l�monqp�rAs�t�u�vxwYy�zK{A|Q}�~���������������A���x�`���=�f�������H�����`�����&������&�` ¢¡&£5¤0¥§¦M¨o©�ª=«�¬H­�®&¯H°�± ²&³#´5µ·¶0¸�¹�ºY»�¼Y½�¾5¿5À�ÁÃÂ�Ä�Å-Æ�ÇHÈfÉ�Ê-ËÌ�Í�Î5Ï0Ð�ÑÓÒqÔ�Õ×ÖKØ0Ù�ÚHÛ#ÜÞÝ�ß�àâá�ãKäæåèç5éMê×ë�ì&í�î�ï�ð�ñ5òôó5õö&÷oø;ù=úAû�üAýÞþ�ÿ������������� ��������������������! "�#%$ &('*)�+�,.-./1032.4653798�:<;>=?�@BA C(D*E�F�G.H.I1J3K.L6M3NPO(Q>R�S�T.UWVXZY\[ ](^*_�`�a.b.c1d3e.f6g3h9i!j\k

l�monqp.rtstu(vxwzyB{}|�~������������3�����t�3�z�������6���3�3�������Z��� �¢¡x£¥¤�¦§�¨3©�ª.«­¬t®(¯Z°z±B²´³�µ�¶B·¹¸�º�»�¼�½z¾B¿�À�Á�Â�Ã3Ä*ÅÇÆ�È�ÉËÊ�Ì*Í�ÎËÏ�Ð�ѹҹÓWÔÕzÖ�×ÙØZÚ%ÛÝÜ.Þtßtà(áxâzãBäæåzçèêé3ë�ìîí�ï�ð(ñòôótõ*ö*÷�ø�ù%ú6ûýüËþ3ÿ������������ ������������������ �"!$#&%('*)+-,/.�0�13254�687�9;:<*=�>@?BA"C�D�EGF�H�IJ�KGL8M8N�OQP�RS�TVU�W�XZY\[8]�^`_ba�c8dfe3gihkj@lkmon5p�qr3s;tbuwvyx�z|{~}G�&�������|�8���������k������/��������@�������5�|������ �¡£¢�¤|¥�¦"§�¨©�ªB«`¬�­8®�¯`°G±�²�³�´¶µG·�¸º¹�»�¼b½¿¾�À�Á�Â�ÃÅÄÇÆ�È�É8Ê�ËGÌ�Í~Î�ÏÑÐ5Ò�Ó�Ô3Õ�Ö@×kØ�Ù�ÚÛ�Ü�ÝÇÞ�ß5à�á�âGã�ä�å�æÇçéè�êìë�í�î�ï�ð|ñ\ò�óõô5ö-÷bøúù(û|üºý�þ�ÿ������������ ������������������� �! #"%$'&�(*)�+�,.- /�021�3547698:<; ='><? @BADC�E7FG<HJI�K.LNMPORQTSVU�WYX�Z\[5]P^*_#`!a7bVcedgf#hjilknmBo�p.q�rjs!tvu�w�x�y{z}|9~g����������������!�}�5�9�P�'�l�����D������������}�9�T�<��¡ V¢�£%¤'¥�¦D§©¨«ª�¬#­¡®!¯°±#²�³�´'µ}¶T·�¸#¹.º«»½¼�¾V¿'À�Á2ÂÄÃ�Å9ÆÈÇ�É%Ê{Ë�Ì�ÍÏÎ<ÐÑ7Ò�ÓnÔNÕ�Ö�×%ØÙ}Ú«ÛVÜ9ÝßÞ9à�á�âã�ä%å%æ}ç è�é5êNë#ìVíîVï9ð ñBò#ó�ô õ�ö{÷�ø�ùÏú<û�ü.ý�þ ÿ���������� �������������������� �"!�#�$&%('*),+.-�/10�2�3"4�57698;: <>=@? A�B.CEDGF�HIKJGLNM�OGPKQSR�TU V>W�XZY�["\�] ^`_"a�bdc�ef gih"j�kml nKodp�qsrZt�u�v�w x>y{z}|�~K��� �i��� ��Z�N�d�������Z�����{���(����d�����N�K�s������ .¡Z¢>£"¤¥§¦©¨�ª¬«�­N®�¯�°>±�²7³"´�µ�¶¸·G¹dºZ»d¼"½;¾"¿�À�Á�ÂÃ�Ä Å�ÆÇ�ȬÉ�Ê�Ë�ÌdͧΠÏÑÐ"Ò�Ó�Ô�ÕiÖ�×§Ø Ù�ÚKÛÝÜ�Þ�ß àâá©ãZä�å æèç7édê�ë�ìÝí�îðïKñGòsóZô�õ�ö÷ùø�ú"ûýü"þ�ÿ���������� ���������������������! #"%$'&)(+*-,/.1032�4+5�6'7�8�93:;�<�=�>@?BADC�E+F�G%HJI�KML�N�O�P�QSR#TVUXW#YZ�[]\�^D_�`badcBe�f gih�jSkml�nJo+p�qr%s�tDu)v�w�xzy�{}|�~'���+����������� �B�����%���S�)������!�������M���)�'�¡ �¢£�¤�¥§¦B¨ª©�«�¬®­)¯±°+²#³µ´�¶%·�¸º¹§»%¼�½+¾�¿�ÀiÁ

 ÃdÄ�ÅÇÆ#È#ÉSÊÌËÎÍÐÏÇÑ�ÒÔÓÖÕØ×XÙÛÚµÜ�ÝÔÞ�ß�àÇáãâÖä�åãæçºè�é ê�ëíì�î�ï�ðòñ�óXô�õ1öÇ÷�øùVúòûýü�þÌÿ���������� �����

������������������! �"�#�$&%�'�(*)+�,�-/.�01�2�34�57698�:<;/=7>@?BADCFEHGDIKJ�LNMPO�QSRUT�V�WBXZY�[�\^]U_a`Ubaced&fhgjiKkHl7monepNqHrest&uHv!w�xDy&zH{}|�~S�����7�a���/�e�P�a���N�N�����h�U�F�}�e�}���<���H���������^ 7¡¢�£�¤�¥§¦�¨*©�ªP«/¬�­�®�¯±°�²K³/´�µP¶/·7¸�¹�ºK»�¼F½H¾D¿HÀKÁUÂÄÃ�Å�ÆÈÇÊÉ<Ë�ÌHÍÏÎÐÒÑNÓÔFÕeÖN×ÙØÒÚUÛHÜHÝFÞ�ßaàUáÒâKã@äKå�æjç/è7éÙê�ëHì�íïî�ðHñNòÙóNô@õ÷ö�øKù@úeûKü7ýKþ&ÿ������������ �������������������� �"!$#&%�')(+*-,�.0/�1)243+5607�8:9<;>=@?BADCE$FHG I0JLK:M�NHO P0Q�RTSVU�WYX[Z]\_^`BacbYd:e�fhgciYjlk�mYnco�pVqsrutwv�x_yz�{�|�}Y~c����]����l�:����B���0���B����������>���B�_���������� �¡£¢�¤�¥c¦�§�¨�©«ª¬®­0¯T°®±0²u³�´@µ¶¸·u¹Hº�»�¼�½¿¾ÁÀ]Â�Ã>Ä]Å�Æ®Ç�ÈsÉ�ÊÌËÎÍYÏ�ÐBÑÓÒ�ÔHÕ×ÖÙØ:ÚÛÜsÝ�Þ�ß&àá�âYã�ä å�æwç�è_éëê�ì@í�î�ï0ð�ñHò]óDô�õDö ÷ø_ù«ú-ûTü�ý�þBÿ������������� �������������������! #"%$'&)(�*,+.-0/�132547638:9!;=<%>'?)@�A�B�CED�F�G�H�I�J�KMLN)O%P�Q'R�SUTWVYX[Z�\^]�_U`�a3b3c�d'e'f7g7h�i�jlknm7o3prqMs%t�u�v%w7x3y{zW|}3~������������������������������������!�7�����.���7��� ��¡7¢�£7¤3¥§¦!¨#©%ª'«)¬�­®%¯[°�±�²�³M´¶µ�·�¸º¹M»½¼!¾M¿�À=ÁMÂ�ÃWÄÆÅ�Ç[È�É�ÊÌËÎÍ!Ï�Ð�Ñ�Ò½ÓÕÔ#Ö�×EØ�Ù�Ú�Û�Ü�Ý�Þß�à�á�â�ãåä�æEçéè�êÆë%ì�í�îðï�ñ�ò§ó�ô7õÕö3÷�øEùYú½ûWü�ý�þ�ÿ���������� �������� ������������� "!$#&%('*),+.-0/214365(798�:<;>=@?�ACB(DFE(G�HJI0KCLM>N�OFP Q�R$S�T U�VXW0YCZ\[F]_^F`badcfehg ikj�lnmpodq�r�sut.v6w,x0yCz|{�}�~>�F�����h����<���d�|�\�����d�������0�h���$�6�����$�F�<�d  ¡£¢�¤ ¥§¦�¨<©@ª�«_¬k­_® ¯(°�±�².³µ´�¶$·h¸C¹uº»<¼¾½�¿6ÀÂÁ�Ã�Ä6Å(Æ Ç�È<ÉËÊ�ÌÎÍ�ÏÑÐ(Ò6Ó_Ô2Õ6Ö<×�Ø�ÙpÚ@Û ÜdÝ�Þàß�áãâä�ådæ�ç@è_é>ê�ë ì(íïî<ðhñFòCó$ô>õ�öh÷6ø�ù<ú�û�üþý�ÿ������������� �������������������� �!�"$#�%�&('$)�*,+.-�/10$2436587:9�;�<�=4>@?BA�CEDGFIHKJGLNMBO�PRQ�SUTIVBW�XZY4[\�]_^G`RaZbRced_f�gih�jlk�menpo�qer�sBtvuRw,xzy�{}|�~��p�i�$�R���������������������B�B����e�e�4���B�}�R�6���� I¡�¢p£�¤�¥�¦¨§B©�ª4«�¬p­�®B¯�°_±³²z´�µ�¶�·�¸6¹�ºB»p¼�½�¾p¿�À�Á

ÂZÃ

Page 29: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Ä�ÅvÆ�Ç�ÈvÉRÊ�ËÍÌ�Î�ÏIгÑ�Ò_ÓeÔ�Õ6ÖØ×�ÙpÚ�Û�ÜÞÝ�ß�àâá$ãåä�æèçié�ê¨ëRì�íBî_ïñð8òôó�õö�÷�ø�ù_úüû�ývþ ÿ������������� ������������������ ��!#"%$'&�(*),+.-0/214356�798�:�;�<>=@?BADCFE�GIH#JLKNMNO%PRQTS#U0VXW4Y�Z\[�]%^N_T`�acb�d�eBfTg�hi�j�k#lm�nporqts#uv#w�xry.z4{|~}��c���������0���4�p�*�R�#�c�#�����%�@���4�r���#����������4 *¡£¢�¤�¥N¦�§T¨*©�ª�«�¬­I®�¯D°²±´³´µ£¶*·N¸,¹�º4»%¼L½¿¾�À*Ác´Ã�Ä'Å´Æ�ÇÈ�ÉTÊ#Ë´Ì�Í�Î�Ï ÐcÑ ÒTÓcÔ9Õ�ÖØ×,Ù�Ú�Û´Ü�Ý�Þ ß4à�á�â ã@ä å*æ%çéè4êë�ì%íTîNïñðñò#ó�ô�õ�ö%÷~ø�ùûú4ü*ýÿþ���������� ���������������������! #"%$'&)(+*-,-.%/02143-5�6879;:<�=?>A@)BDCFEHGJI�K�LNM#O�PRQ�SUTHV�W+XJY-Z#[%\?]_^�`�acb�d-e�f�gHhji�k�l mn%o�p%qFr#s�t�uvxw�yFz|{�})~;�x�-�����2���#�?�����;�N���������������-������������?���)  ¡%¢4£

¤U¥§¦ ¨ª©?«¬�­j®°¯²±;³;´�µ·¶¹¸º¼»�½�¾·¿;À�Á·Â;ÃÅÄJÆ�Ç;È°É°Ê4Ë�Ì°ÍÎ'ÏÑÐjÒÔÓ?Õ?ÖØ×NÙ!Ú�ÛHÜ�ÝßÞàâá%ã%ä?å'æèçêé!ëHìîí!ï�ð�ñ�ò%ó�ô)õ ö?÷ùø-ú%û!üjýþ_ÿ������������ ��������������������� �"!$#&%�')(�*,+-&.�/1032�4�5,687:9$;=< >�?A@�BDC$EF�G�H�I�J�KL�M�NPO=QSRUT�V W�X Y�Z�[U\8] ^�_S`Ua8bDc�d e�f�g�h�i jlknm�onprqtsvuxwzyl{�|}r~l���S�"�z�,����� ���&�U���������=� ����r���,����¡  ¢r£:¤�¥P¦�§ ¨ª©�«­¬�® ¯°:±�²¡³ ´�µª¶�·,¸º¹�»�¼�½S¾�¿ÁÀ�ÂrÃÅÄ�Æ�ÇUÈ�É�Ê ËrÌlÍÏÎ�Ð�ÑÓÒ�Ô Õ�Ö=×ÙØvÚxÛzÜlÝ�Þßà�á�â ã�ä=åUæ�ç,è­é�êëíìrîlï�ð�ñvò�óUô&õ=öø÷�ù�úUû"ü�ý�þ ÿ ��������� ��������������������! #"�$&%('*)&+�,.-0/21 3�4�57698;:=<?>7@BA�C�D.EGF�HJI�K�L�M�NO�P=QRTS#U�VXWZY\[^]�_a`cbZdfe�g0h#i�jlknmpo�q&r#sctfu&vxw#y9z|{G}&~������������T�=������9�f������� ���#�7���;�������0����� �¢¡¤£0¥�¦¨§=©9ª0«#¬#­¯®�°²±�³µ´�¶X·¸�¹�º!»\¼¢½=¾(¿=ÀÂÁGÃ�ÄÆÅ�ÇÈ�É(Ê�Ë�Ì!ÍÎ?Ï�ÐÒÑXÓ9Ô�Õ;Ö�× ØTÙ#ÚpÛ�Ü.ÝcÞ�ßJà\á�â¢ãcäXå�æ=ç*è�é�ê\ë0ì�í�î�ï ðTñ0ò&ó#ôJõZö�÷

ø#ùú�û�üTý þ¢ÿ ���������� ����� ����� �����������! #"%$�&�')(�*,+-�.0/21�3�4 5�6�7�8�90:�;=<?>A@�BDC�E�F�GIH�JAKML�N�OQP�RTSVU�W%XZY\[^]#_!`�acb�d�efMg�hji2kml�n oqp!rtsuwv�xjy{z|~}��A�������=����#���t���������=�j����{�Z�����j�#����2�M�!� �¡£¢¤2¥^¦?§{¨�©�ª�«­¬¯®�°�±!²´³�µ�¶%·�¸�¹»º�¼A½´¾�¿2ÀDÁDÂ�Ã�ÄcÅ�Æ´ÇmÈ�É�ÊÌËMÍ�Î�Ï�ÐÑ?Ò{ÓMÔ�Õ�ÖÌׯØ�Ù�Ú=ÛÝÜ�Þ�ßáà�â�ãtäjåcæ�ç�è�é�êìëí#î�ïMð�ñ%ò�ó2ô�õmöø÷�ù�ú´û�ü#ý�þ�ÿ������������ ��������������������� "!�#"$�%'&)(�*,+.-"/102�3547698;:�<.=?>'@7A�B�C�DE�FHGIKJ�L�M"N;OQP�R�S�T"UWVYX[Z]\�^"_�`�a�bdcfe�g[hjiHkml�n�oqpsrut9v,w�xy)z{f|�};~��"���[���?���?���,���j���������7�)�.�"���.�?����������?�� �¡.¢,£¥¤§¦�¨©�ª7«7¬�­,®'¯f°H±"²´³fµj¶�·�¸º¹¼»¾½"¿jÀÂÁ"ÃÅÄ�Æ�ÇÉÈjÊ�ËÍÌfÎ7ÏÑÐ�ÒjÓ�Ô[ÕÖ�×"Ø�Ù§Ú)Û�Ü�Ý�Þ�ß�à�áQâHãåä�æjçéè�ê)ë�ìîíjï§ðjñ"ò'óõô,öø÷§ù�úéû)ü�ý7þ�ÿ�������������� ����������������������� "!$#"%$&�'�(�)�*,+�-/.1032�465

7/8:9$;�<�=�>@?BA�C�D:E:F�GIHKJML�NPORQTS3UKVXWZY\[^]�_�`badcfehgji�khl3m�nXo:pq�r�sXt"uTvZw@x�y$z|{�}�~����Z�����|�j�����Z���P���B�Z���B�1���P���K�����P���^���� ¡Z¢P£�¤�¥X¦Z§�¨X©"ª¬«®­°¯3±B²�³µ´�¶�·�¸,¹:ºT»X¼d½�¾\¿,À"Á"ÂZÃÅÄ�Æ�Ç,È�É�Ê�ËÍÌhÎ:ÏZÐ�Ñ�Ò�ÓÔÖÕ�×"Ø�ÙPÚ$Û3Ü�Ý�Þ6ß6àKá�â/ã�ä�å�æKç|èIé$êPë,ìÅíBîZï@ð�ñ�òôó�õ÷ö/øZù�ú�û®ü/ý�þ�ÿ����

������� ���������������������! #"%$'&)(!*'+-,/.�02143�57698;:;<;=?>�@ACB4D?E�FHG7I4JKML'N;O9P4Q7R;S%T�UCV�W#X2Y�Z\[^]_�`\a�bced�fCg�hji2k%l�monqprCs�t�ujv�w�x�y\z7{|~}'�#�?�����?�;����#�!���q���M�������7�2���!�j�-�������H �¡�¢�£M¤�¥�¦�§C¨ª©�«7¬­'®ª¯C°²±´³ µ·¶¹¸-º)»%¼C½o¾�¿�ÀeÁjÂ;ÃeÄÆÅ�Ç�ÈeÉ�Ê�Ë�ÌÍÏÎ;Ð�Ñ;Ò2ÓÔÕ×Ö2Ø%Ù#ÚÛ Ü)ÝÞ%ß7à´á4â'ã�äjåçæHè/é)ê2ë2ìîíÆï�ð�ñ!ò!ó�ôjõ�ö²÷�øÏù^ú?ûüCýÿþ������������� ���������������������! #"%$'&)(�*,+.-�/�0214365�7,8�9;:<>=!?6@�A�B�CEDGFIH�JLK�M!NPORQ>S�TUWV�X6Y6Z![E\�]�^�_P`�abdcfehgGi4j�k!l�m4n)o�p�qsr4t�uwv�xzy6{�|�}�~����P�#���4�6�!��E��2�������6���������)���w�������%� �¡)¢6£¥¤,¦'§�¨!©�ª�«W¬�­>®�¯�°�±#²R³�´Rµw¶!·R¸�¹)º!»¼¾½À¿WÁ�Â!Ã�Ä�ÅEÆ#Ç�ÈÊÉ)ËÀÌ�Í!Î�Ï�Ð�Ñ4Ò�ÓwÔ)Õ×Ö�Ø�Ù>Ú ÛIÜ�Ý�Þ�ß2à4á6âsãWä4å6æ�çèêé�ë6ì�í�îPï�ðGñ,ò)ó�ô!õ!öê÷�øúù6û)üþý ÿ ��������� ������������������� �! �"$#&%('*)+�,.-0/21$3(46587:9�;=<?>�@(A�B�C$D�E�FHG$IKJML!N OQPSR(T�U�VHW XZY�[!\](^$_a`�bdc�egfih2jgk laminpo�qMr(s�tMuwv=xzy8{*|Q}�~p�����������*�p���a�����$�$�w�(�*��������g� �������=���H (¡�¢Q£(¤¦¥!§�¨�©!ª!«2¬g­!®°¯ ±�²´³¶µ�·�¸¶¹$º¼»�½H¾�¿ÁÀHÂ�Ã�Ä ÅgÆ�ÇÈ�ÉËÊÍÌÎ�ÏÑÐ$ÒÔÓ�Õ�Ö�×�Ø(Ù�Ú�ÛHÜ*ÝwÞ=ßià=á(âgã�ä�åçæ0è2é�ê2ë�ì2í0î�ï�ðñ

òÑógô�õ�öQ÷ øÍùûúgü�ý*þ�ÿ������������� ������������������ ��!#"%$�&�')(+*-,/.10#23�4)576�8/9;:=<�>@?�A�B�C�D�EGF�H�I�J�KML�N�OQPSR�T�U;VQWYX[Z]\_^/`ba)c[dfe�g�h�ij)k�lbm)npo#q�r�st/u�vxwzy1{}|�~��)�Y�x�%�S�����)���-�����[�p�������#�=��������%�x�/�� �¡=¢¤£=¥x¦¨§ª©�«/¬�­#®Y¯=°�±x²[³b´¶µ=·¹¸=º�»�¼�½¿¾�À�ÁG¹ÃxÄÆÅbÇÉÈ�ÊË�Ì�Í�Î�Ï�ÐSÑ[Ò�Ó�Ô�Õ�ÖØ×�Ù�Ú�Û1ÜÞÝxßáà[âØãåä/æèç�é�ê�ë�ì�í�î�ï�ð�ñ)òôóxõ÷öèø=ù_úYû�üýÿþ���������� ���������������������� �"! #%$�&�'�(*),+.-0/�132�4�57698;: <>=? @ ACBD,E�F G>H*I J�K�LNM�O�P�Q"R�S�T�UWV�XZY\[�] ^`_�a b9c9dfe�g"h�i�j k lnmpo9qr�stvuxw�y*z|{�}�~��*�������9����������n���;�Z�����*����� �����9���% Z¡v¢*£0¤�¥�¦�§�¨>©�ª¬«�­®0¯�°;±³²�´�µ¶ ·¹¸�º|»�¼ ½Z¾ ¿�À�Á�Â*Ã�Ä>ÅÇÆ È,É�Ê�Ë Ì0Í�ÎCÏ*гÑ0Ò Ó Ô0Õ�Ö*×ÙØÇÚ�Û7Ü>ÝÞ"ß àâáZã�ä0åÙæ%ç�èêéÇë9ì í%î*ïnð"ñ�ò�ó�ô õ*ö*÷ øúù�û üþývÿ���������� ���������������������� �"!�#%$'&�(*)�+-,*.�/�0'1�2%3"46587:9<;>=<?A@�B�C-D8EGF�H�IKJ�L�M�N-OQPSRTVU�W�X�Y�ZA[�\"]>^�_�`ba�ced�f�gih6j*k�l�monqp*r�sitGu�vAw�x�y�z {Q|~}6�*���������K�q������� ��e�����������8�i�8����� ���b���8 �¡i¢¤£�¥<¦�§�¨�©Gª8«q¬®­i¯±°�²�³µ´8¶>·A¸*¹~º»¼<½�¾>¿�À�Á�ÂÄÃ8ŤÆ�Ç:ÈKÉËÊKÌ�ÍeÎ�Ï�ÐÒÑQÓÕÔiÖ�פØ�ÙÛÚ�Ü�Ý®ÞQß6àQá6â-ã8ä�å�æ�ç�èËéQê'ëì�í�îqï�ðqñÄò�óQôQõ*öø÷�ù8ú ûbüøý*þ ÿ �������� ��� ������������������� �"!$#&%('*)�+,�-/. 0�132547698;:=<">*?A@5BDCFE�GIHKJ�LNMPO�QKRTSVUXW�Y�Z[]\�^A_�`$aDbVc�d=egf hji(kl�m"n�oqp�rts�uIvxwqy�z|{�}�~��|�����(����� �N�����]������A�����5�V���������I���� �¡N¢�£q¤�¥]¦¨§A©Iª�«D¬V­�®°¯�±�²´³ µ"¶$·&¸(¹*º�»¼�½¿¾;À ÁÂ�Ã7ÄDÅ7Æ$ÇDÈ ÉËÊ�ÌÎÍDÏ�Ð7Ñ=ÒÎÓAÔ�Õ]ÖD×qØKÙKÚjÛ�Ü�ÝjÞ�ß�àâá]ã�ä�å�æVç;è�é�ê�ë ì"íqî&ï(ð*ñóòôDõ÷öùøDúDû÷üqý�þ ÿ ��������� ���������������������! #"%$�&('*),+.-0/213�4*576989:<;(=�>@?�A%BDCFEHG�IKJMLON9P�QHR<SUTWV�X<Y[Z*\*]_^�`ba�c@dFe2f�gHh�ikj�lnm.oqpr�sut�vHwyx�z|{�}y~��W���W�#���������*�[�H�%�%�W���(�����H���%���9�[�K�@�7�0 *¡¢£K¤H¥�¦0§�¨�©«ª7¬®­�¯H°²±7³�´Oµ0¶<·�¸¹�º²»�¼W½(¾*¿�ÀÂÁ9Ã�ÄqÅ[Æ�ÇÉÈOÊ(ËÍÌqÎÐÏHÑ2ÒÓ�Ô*Õ�Ö.×MØ#Ù7Ú�ÛOÜuÝFÞ7ß�àâá7ãHä�å#æ�ç9èué�êìë�íHîðï�ñ|ò�óyô�õ<ön÷DøFùHú�ûýü�þ@ÿ����������� �������������������� "!$#&%')(+*-,�.0/�132�4�57698;:�<$=?>A@B�CDFE0GAH7I;J)KMLONOP;QSR+T�U�VXW0Y[Z�\�];^�_�`)acbdfe;g�hjilk&m3npo�q�rtsvu�wx;yz�{"|[})~������0�����������M���[�$�-�;�����l��������� �-��� �)¡�¢3£¥¤;¦§7¨[©lª7«¬®­-¯"°± ²�³´-µ-¶®·¹¸»º�¼ ½M¾�¿ÁÀ)ÂAÃ�Ä7Å�Æ�Ç)È É�Ê3ËSÌMÍ7Î ÏÐÑ�Ò�Ó�Ô�Õ�Ö×�ØÚÙ)ÛMÜÚÝ�Þ�ß)à0á�â0ãlä7åçæ�è�éëêíìïîSð�ñ+òló&ô3õÚö�÷$ø7ùÁú�ûFüïý�þÿ������������ ��������������������! #"!$&%('*),+�-&.�/10&2!3�46587,9�:;�<,=�>@?BADCFE(G,H!I�J(KMLONQP(RTS�U�VXWZYX[Z\,]_^a`cb�dfe�gFhji�kjlTmMn�o!p�q!rsut#vw,xzyM{�|�}#~�O�*���j�z�!���Z�!�@���Z�������!�����z�����f���#��j�T *¡�¢

£¥¤§¦#¨�©ª¬«c­�®°¯#±³²�´&µ�¶&·�¸#¹cº¼»¾½�¿8ÀÂÁÃ�ÄQžÆZÇ�È�É�Ê8Ë�Ì�ÍÏÎ(Ð8ѬÒ*Ó�ÔÖÕØ×*ÙÚ#ÛjÜ�Ý�Þ*ß�à¬áÂâfãåä�æ#ç�è&écêÂë¬ì#í�îðï¾ñ�òðó,ô�õ#öu÷ùøDúcû�üfý�þ&ÿ���������� ������������������ �"!$#�%�&('*)�+(,�-�.�/10�2�354(687$9;:=<�>@?BA�C=D�EGF8H5I(J8K$L8MON�P=Q�R SUT�V�W�XZY([\$](^�_(`�a$b�ced"fhg�i j�kmlUn�o8p qUr�s8t8uwv�x=y�z8{�|=}"~;�U���(���

�����B���=���5�8���Z�"���G�Z�B�*�=�B���� ¢¡¤£B¥ ¦ §©¨*ª(«�¬®­°¯�±²�³(´�µ=¶�·B¸G¹»º(¼;½

¾�¿

Page 30: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

À=Á�ÂÄÃ�ÅÆ"Ç©È�ÉGÊ�ËÍÌ�Î�ÏÑÐ�Ò*Ó�Ô�Õ�Ö5×�Ø�Ù�ÚÜÛÞÝàßÞá�âGã�ä"å©æàç�èBéëê"ì*í�îðïñÄòGó�ôÄõ;öU÷�ø�ùûúÄüGý�þ©ÿ���������� ��������������������� �"!�#%$

&')(%*,+.-0/"132�46587:9<;�=:>�?A@CB�D�EGF�HJI"K�L�M�N O)PRQCS�TVU�WYX�Z\[]�^ _"`Va b"c<dfe,g�hVi)j�kmlonqp�r0s:t.uwv<x<y0z"{"|} ~w�6���������6�"�V�)�.�����C�����,���C��,�����)�<��� �V¡ ¢�£o¤¦¥ §0¨ª©�«8¬:­�® ¯±°f²6³q´�µ·¶�¸C¹�º:»�¼V½�¾À¿"Á Â�ÃÅÄ�ÆÇ È:É�Ê�ËÍÌ Î0ÏÑÐCÒ<ÓfÔYÕ±Ö,×6ØoÙ<ÚÜÛ�ݱÞoß)à%áãâ�ä�å�æ,çVè�éCê<ë·ì,íwî,ï�ð ñóòô�õ�ö6÷6ø�ù±ú0û.ü�ýVþ ÿ ��������� ������������������� �"!�#�$&%')(+*-,/.10�2�35476�8�9;:�<=?>�@"A"B�C�DFE�G�H7IKJML7N�OQP&RTSQUWV�X�Y�Z�[T\ ]"^`_�acb�d�egfKh�i�j;k�l m�n`o�prq`s?tu�v&w?xgy�z|{�}�~����Q�`�����&�M�������?���&���)� �"��� �g� �W������� �7�� c¡�¢+£r¤�¥¦�§g¨r©�ª|«7¬®­r¯`°-± ²�³ ´�µ·¶W¸�¹»º ¼"½�¾�¿ÁÀgÂTÃMÄÆÅÇ)È7É�Ê+Ë Ì Í7Î|Ï Ð&Ñ®ÒÔÓÕÆÖ�×·Ø ÙrÚ`Û�Ü5Ý�Þ7ß�à�á�â»ã�ä�å�æ"ç)è+éëê�ìîí·ï7ð)ñëò-ó ôöõ�÷�ø»ù ú�ûTü�ý þ-ÿ������������ ��������������������! #"%$�&('�)�*�+-,�.0/21�3�46587:9<;�=%>#?�@BADC�EGF�H�IJ%K!LNM8O2PRQ�S:T:U0VXW�Y�Z�[]\�^<_`a�bdc#egf�h�i�j�kml%npo!q�rts8uDv<w%x�y#z�{|<}D~!���D�0���N������0��������!�������%�6�����0���������� �0¡!¢¤£:¥<¦�§0¨-©�ª¬«<­:®�¯

°�±!²�³�´ µ·¶�¸#¹»º8¼�½0¾�¿�À6Á0ÂDÃ0Ä#Å�ÆÈÇ(ɻʷË<ÌÎÍ�Ï%Ð�Ñ�Ò�ÓÕÔ�Ö�×ÎØ�Ù#Ú»Û»Ü#Ý!Þ�ßáàâäã�å8æ�çéè�êmë%ì�í î�ï#ð»ñ0ò!ó0ô�õ�ö�÷<øùÈúüû<ý�þ ÿ���������� �������������������� �! #"%$'&)(+*�,.-0/�13254 6�7 8:9:; <>=@?BA�C�D�EGFIHKJ�LNMIOQPSR#T!U:VWNX�Y3Z5[�\^]`_:acbed#f@g�hjiekNlnmpoIqsr!tKu#vSw#xSy�ze{}|�~@�S�:���!��� �0������>�����:���+�S�#�)�>���Q���j������� �¡ £¢¤�¥N¦>§�¨ ©:ª£«�¬Q­�®�¯ °�±³²#´#µ!¶N·�¸¹�º�»�¼�½N¾�¿jÀ�Á'Â}Ã�ÄÆÅ:Ç�È�ÉËÊ>Ì�Í#ÎÐÏ�Ñ^Ò^Ó#Ô@Õ×Ö�ØeÙQÚÜÛ+Ý�ÞNß�à:áâ�ã�ä�å!æ�ç�è�é ê�ëQì�íïî�ð�ñ5ò�óeôöõs÷!ø�ù�ú�û>ü�ý�þ#ÿ���������� �������������������� �"!$#%�&('�)+*-,/.1032�4+57698":�;=<?>$@BA�CEDGFIH JLK�M9NPORQ S+TGU1VIWX�YBZ?[�\]7^`_acbBd"e�fhg�i1jBkmlonqp+rEs1tIu?v�w xzy"{}|�~���B�����

�h���3���`��m�������������=�$������"�/�B������ 1¡�¢o£¥¤�¦I§�¨o©�ª�«o¬�­}®�¯±°²³$´9µ ¶1·"¸º¹�»}¼m½±¾�¿}À�Á�Â1ÃÅÄ3Æ Ç�ȱÉ�ÊBËcÌ"ÍhÎ}Ï�Ð�ѺÒmÓ Ô}Õ3Ö�×BØ=Ù�Ú$ÛÜBÝ¥Þ�ßBàá�â$ã�ä�åBæGç1è�é`êBëíì�î�ï$ð ñ1òEó�ô�õmöø÷"ù�ú"û�ü�ý�þEÿ������������� ������������������ "!$#&%('�)+*-,�.0/�1�23540687:9<;�=?>�@-ABDCEGFIHKJ�L5MON0P�Q�R0SUT�VXWZY[]\"^`_�acb�dfeOgXhZi�j�k�lDmDn0o�p�qDr�sXtvu�w�xvy�z|{`}D~����<�5�<�0�����

� �f���c���c����|���`�c�`������� ��¡�¢ £�¤�¥§¦©¨�ª|«�¬`­®`¯c°²±³²´ µc¶�·c¸�¹�º»(¼�½&¾�¿�ÀÂÁcòÄ

ÅÇÆUÈ$ÉËÊUÌ�Í"Î�Ï]ÐÒÑ�Ó+Ô�Õ�Ö×UØ$Ù ÚDÛ`ÜÝIÞGß�àâáãåäæ�ç�è"é"ê�ëKìåíUîðïvñ�ò|ó<ô�õ|ö÷Dø�ù&ú�û�ü<ý�þDÿ���������� ���������������������� "!$#$%�&('*),+.-$/�02143�57698:;=<�>*?�@BA9C�DFEHGJI�KLNM2OQPSR=T�U*VXW�Y Z�[$\^]`_�aQb$ced2fHg7h9iHj"kHlnmporq s"tuv�w�xzy{7|~}�����~�2�$�����.�.�n���.�����B�����9�J����������� �"¡¢.£2¤¦¥4§�¨�©Hª"«H¬�­=®r¯°"±�²�³ ´pµ�¶9·"¸�¹.º"»n¼9½7¾�¿�À ÁÂJÃNÄÆÅ�Ç�È*É�Ê�Ë Ì`Í�Î"ÏÐѦÒ�Ó=ÔÖÕ�×�ئÙ4Ú�Û9ÜHÝ7Þ�ß�à�áHâ�ã�ä å æ=ç.èJé�ê�ënì"í�îHï7ð�ñòôóJõ÷ö�ø ùpú�û9ü"ýÿþ������������� ��������������������! #"�$�%'&�(*)�+�,

-�.0/�1*2�3�4�56�798;:=<�>�?A@�B�CED�F0G�H!IKJ LNMPO�QSR�T9U!VXWZY0[]\]^S_�`a0b�c�dXegf�hjilknm9oqplr�sPtvuxwNy�z|{~}X���K�����N���9�S�~�N�q�X�x��������������~�N�9������� ��¡£¢¥¤§¦�¨9©!ª�«­¬�®�¯±°q²�³µ´P¶]·¹¸Sº�»�¼�½�¾~¿ÁÀÃÂ�ÄÅ�Æ0Ç�È­ÉËÊN̹Í�Î�ÏPÐ�Ñ�Ò~Ó�Ô�Õ�Ö'×SØKÙ�ÚÜÛ�ÝjÞàß'áÃâNã�äPå�æ±çéèëê�ìNí�îXï­ðñgò­ó�ô�õÜöø÷úùüû�ý­þXÿ������������� ������������������! #"%$'&)(+*,�-/.1032541687�9;:/<>=+?A@CBEDGFIHKJ#L�M)N'OQPSR�TVU�WYX[Z]\�^�_a`cb/d�e�fhg

i;j�kSlCmon�pGqsr#tEuEvCwyx/z){�|G}h~1���c�3�I�����'�o�;�������5�1�y�a�o�)����E�������'���a�)�3 �¡'¢�£ ¤�¥S¦E§S¨�©5ª;«�¬Y­ ®�¯�°�±�²E³Q´'µ�¶ ·�¸]¹�º�»�¼h½V¾¿5À�Á'Â�ÃÅÄ�ÆCÇ�È�É]Ê5ËÍÌ�ÎSÏ�Ð�ÑEÒ]Ó5Ô�Õ5Öh×)ØÚÙÅÛÝÜ�Þyß�àsá�â�ã�äIå�æ�çEèVéEê+ë�ìîí�ï�ðñ]ò)óEô�õ÷ö]øúù]ûEü�ýsþ�ÿ���������� ��������������������� �!#"%$�&')(+*-,/.�01�2/3%4656798�:�;/<>=?A@#B9C�D�E FHG�IKJ#L)M NPO QSR�T�UVHWYX�Z\[^]�_�`�a�b

c�dfegih�jlk/m�n�oqp�r#s%t#ufvxwly{z}|�~��9�l�/�l�/�P�����>���P���������������� ����������Y �¡�¢%£9¤9¥Y¦l§©¨�ª%«�¬^­�®°¯l±�²�³/´%µ#¶�·�¸#¹Pº�»P¼¾½�¿�À©Á%Â�Ã�Ä>Å�ÆYÇ�È�É�Ê�Ë Ì�Í�ÎÏ�ÐPÑ�ÒÔÓ�ÕiÖ�×%ØÚÙ�ÛiÜ ÝYÞ�ßáàxâYã�äæåiçxèêé ë/ìîí)ï�ð{ñóòôPõ÷ö�øù�ú�û)ü#ý�þÔÿ����������� ����������� ������������� ���! #" $&%('!)+*-,!.0/(132546�798;:=<#>#?-@A5B;CEDGF0H5I;JLK(M&NPO&QSR+TVU!W�X�Y[Z]\�^#_�`Ea

bdcfehg+i�j-k!l#monqpsr tvu�wfx+y�z]{&|3}3~S��(���=���V�(�#������-�����h�s�� �E�q���q�������]�����3�v ;¡£¢#¤5¥ ¦¨§+©]ª&«­¬f®;¯s°­±9²3³]´�µ�¶3·�¸º¹¼»¾½¿ÁÀÃÂ�Ä3ÅÆ;Ç=È;É5ÊfËÁÌqÍ]Î3Ï Ð5Ñ�Ò�Ó+Ô]Õ(ÖØ×�Ù#Ú+Û!Ü(Ý;Þ�ß�à á�âfã�ä+å�æ+ç�èêé ë ìÁíî0ï5ð�ñ�ò]ó3ô õ öS÷]ø�ù ú û¼ü�ý�þ ÿ������ ������ �������������������� "!$#&%('*),+�-/.�021�354�687:9<;�=?>A@CBED�FCGAHAI<JLKNMPOQ<RTS�UAVXWY[Z�\^]`_�abdcfe<gih�j�k�l�m�npoPq�rts�u�vdw,xiyzp{i|P}�~�<���p���������C�p������P�E�8���,�������P�����X�� �¡�¢�£�¤<¥i¦C§�¨i©Pª�«�¬®­°¯,±T²i³,´�µ�¶P·¸�¹�ºP»½¼�¾p¿pÀÂÁXÃ�ÄPÅÇÆ<ÈiÉCÊpËiÌCÍiνÏ(ÐAÑ�ÒCÓiÔ�ÕCÖ�×�ØÇÙ�Ú8ÛCÜ�ÝßÞià�á°â�ãåäßæçè�é8ê�ë�ì�íî<ï&ðAñ�ò8ó�ô�õCöå÷�ø�ù�úpûüCý�þ�ÿ������������� ������������������ ��!#"$�%'&)(+*-,�.0/ 1#2�3545687:9�;=<�>�?-@BA�CED�F�G�HJI K�L+MON P#Q�RTS5U+VXW�Y+Z[�\�])^`_=a�bdcdegfihBj�kmlongp-qsrutwvyx=z {�|B}�~X���B�:���������������������X�d����������8��������� ��¡£¢�¤�¥�¦�§�¨ª©�«�¬�­-®�¯±°�²�³µ´+¶¸·)¹8ºX»`¼�½m¾À¿�Á�Â�Ã=Ä�ÅÇÆ�ÈÉ+ʱË�ÌdÍ�ÎyÏ�ÐÒÑÔÓdÕ)ÖØ×-Ù-ÚEÛÝÜ�Þ=ß�à�ásâ�ã=äæåXç èêé�ëyì�í�î�ïñð�ò�ó�ôöõ÷�ø=ù ú�û�üÒý�þ`ÿ���������� ������������������� ��!�"$#&%('*),+.-�/�01�2�354687�9�:�;=<?>�@�A�BDCFEHGJI�KMLON(PRQTS�URVJWX�Y�Z\[�]_^�`�acb.d�eHf g�hikjml.n?o�p5qOrsJt,uv,w�xzy�{=|J}H~H�J���M�O���R�*��������������J��� ����� �������O�=�D�F m¡m¢�£¤.¥�¦�§�¨ª©¬«�­T®¯ °²±O³�´Oµ�¶,·¹¸»ºm¼5½�¾.¿OÀ�Á ÂÄÃÆÅ5ÇzÈÊɪËÄÌ�ÍHÎ.ÏzÐÑ�Ò�ÓÔ�Õ»Ö�×mØ?ÙÛÚ�Ü_Ý=Þ,ß�àâá�ãkäzå.æTçzè5ézê ë�ì�í_î�ï.ðñJò.ó�ôªõ?öF÷�ø�ù.ú�û�ü ýþÿ����������

� ������������������� �"!$#&%('*),+�-/.10"2 354�687:9<;>=@?BA,CED F:GIH�J1K"LMONQP:R<S�TVUXW�YIZ [�\^]`_�a�bdc e5f/gih�j�kml�n�o:p<q�r>s�tvu^w/xzy/{}|�~5�`�z���Q��/�X�/�X�^�"�����`���I������������/�`�"���<���, z¡`¢�£"¤v¥¦"§�¨�©`ª�«­¬`®O¯Q°�±v²´³µ^¶z·�¸>¹"º »�¼^½¿¾IÀXÁ>Â�ÃOÄXÅ�ÆÈÇÊÉÈË�Ì�Í�ÎzÏ�ÐQÑÓÒ�ÔOÕ�Ö,×"Ø"ÙÓÚ/Û/Ü�ÝßÞ/à�á<âã äQåæ"çéè`ê"ëIì,í�î5ï�ðßñ�ò�ó@ô�õ(öv÷�ø$ùÈú�ûýü"þ�ÿ���������� ��������������������! �"!#%$�&�')(�*,+.-�/10�2�3�465�7)8�9,:<;>=6?A@CB�DFE1G�H,IKJ!L�M�N�OFP�Q�RSUTWVYX[Z�\W]

^ _a`cbed�fhgei!j�kcl

monqp�r1s�t<u�vWw�x�y{z}|�~U���������������������!����������,���h��������)�������,�U ¢¡[£¥¤§¦a¨�©!ª�«!¬%­�®�¯)°�±,²´³�µ�¶·¹¸Fº�»½¼[¾�¿½À�Á�ÂÄÃÆŹÇYÈ[É�Ê�ËÌ[ÍUÎ[Ï%Ð!ÑÄÒWÓÕÔAÖØ×UÙ�Ú�ÛÝܹÞ�ߥà�á¥âäã�åUæ�ç�è!é�êUë[ìKíWî,ïeð%ñ�ò�ó%ô�õ�öø÷�ùú�û�üþý¥ÿ���������� ��������������������! #"%$'&)(�*!+-,/.10!243/57698;:#<>=@?9ACBD E F�GIH�JK7LNMPO/Q�RTSVU9WCXZY�[\�]9^#_)`ba�ced�fPg�hTikjTlkm�nporq�s t7uwvpxyCz{9|N}~����C�������!�w����7����� �9�r���N���T�!���P���9�C�k���;�¡ 9¢7£>¤¦¥N§P¨7©�ª«P¬�­¯®�°7±!²w³w´Cµk¶�·�¸P¹Cº9»½¼%¾p¿�À�ÁÂ�Ã�ÄÆÅ1Ç�È�ÉCÊ�Ë�ÌCÍeÎ7Ï�Ð�ÑpÒrÓ Ô9Õ;Ö!×�Ø�ÙÚTÛÝÜIÞ�ßCà)á7â/ã/ä�å;æèçTékêTëkì�ípî7ï�ðbñ�ò¡óõôröp÷rø)ùûúIürýÿþ������������ ������������������ �"!$#&%('*),+.-"/�0214365*748�9;:6<>=�?A@6B�C;DFE4G2HJILKJMN�OQPSR�T*U�V4WYX�Z\[^]Q_S`badcfehg�i�j�k4l�mon�prq4s�t&u�v�w�x�ySz�{�|�}r~6�b�6�2����,�f���J�����������*���*�J���o���������f���2 �¡2¢¤£S¥�¦�§S¨F©�ª¬«*­¯®^°Q±S²b³d´fµ¶�·4¸�¹4º�»F¼�½r¾S¿�À>Á2Â�Ã�Ä�Å\ÆJÇ�ȤÉ^ÊÌËJÍbÎdÏfÐ$ÑÓÒoÔLÕdÖJ×6Ø�Ù�ÚÜÛÝrÞYßdà�á�â,ãåäLæ�ç6è*é�ê^ë�ì�í�î*ï�ðòñ�ó4ô�õ4ö*÷oø>ùrúSû�ü�ýSþ2ÿ���������� ���

�������������������! #"%$'&)(�* +,�-�.�/�021�35476�8:9�;�<=?>�@!ACBED�F�G�HJI�K

L�MON:P�Q!R#SUTWVYX�Z7[]\_^a`7bCc_d!egf7hji%k l mgnpo7qsr!t7uWvWwyx]z!{:|~}j���������7���� �g���g�g���������p�g�~�s�������7�?�����g ~¡�¢�£�¤�¥!¦g§~¨y©!ª «7¬p­%®°¯%±�²?³?´

µ¶

Page 31: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

·�¸�¹#º»�¼�½'¾�¿_ÀWÁWÂ�ÃÅÄ�Æ�ÇÉÈËÊÍÌÏÎWÐOÑ�Ò!Ó]Ô�ÕÍÖC×ÙØ?ÚÜÛWÝJÞ�ß�à!áãâ�ä�åæ�çè%é�êjë_ì�í�îÙï�ðòñ�ó�ô�õ÷ö�ø�ù�úüû�ý�þCÿ���������� �������������������! #"%$'&(*)�+-,/.1032!435�6879!:<;>=@?BA*CED@F�GIHKJ>LIMON�P�QSR�T*U�VXWZY*[�\�]_^'`�ab�ced�f�g�h�ikjel*monqpsrutwv#xzy�{}|B~������B�*�����e�#���e�}�u���������O�1��*�'�����X�'�X�� �¡£¢!¤'¥�¦e§*¨o©«ªO¬'­z®1¯±°>²!³µ´O¶#·�¸X¹�º¼»�½#¾!¿�ÀSÁÃÂ�ÄSÅÆZÇ£È�É8Ê�ËeÌÎÍSÏeÐÎÑ�Ò�Ó!ÔeÕIÖØ×eÙ!ÚIÛ}Ü#Ý/Þ�ßeà!áãâØäeåçæ�è�é�ê>ë>ìÎí�î�ïçðqñ>ò!ó�ôõ*öø÷�ù3ú�ûeüSý£þOÿ���������� ������������������������ "!$#�%�&�'�(�)+*-,/.0�1325476+8-9�:-;=<?>/@BADC�EGF�HJILK�M�N/OQP"R�STQUWVYX[Z]\�^]_�`ba"c�de"f/g�h�ikj�lnmYo5pnq+r�s7tBuwvyx{z |~}��y�B���=�����/���k���~�Y���������-���-� ���~����7 7¡�¢¤£�¥-¦"§©¨�ª�«�¬�­L® ¯°{±³²´¶µ"·�¸©¹�º3»{¼�½¾�¿ À�Á"ÂÄÃ�Å�ÆÇÉÈËÊ7ÌÎÍyÏÑÐ�Ò"Ó�Ô~Õ�Ö=×�Ø"Ù�ÚÜÛ/ݤÞ�ß~àÉá[â+ãåä�æ�ç[è]é"êìëíïî~ð�ñYò[óõô�ö+÷Yø[ù{úüû+ýYþyÿ������������ ������������������� ��!#"%$�&�')( *+-,/.0�1323465�798�:/;=<?>�@�ACBD-E3FHG?IKJ�LNM6OQP�RTSVUCWHXTY[Z]\-^�_a`�bCc�d�eVfhg i[j kTlnm9o�pqsrut�v�wyx{z}|�~?�-���3�u�/���}�������������-���������� �?���T�?�{���  ¢¡)£C¤�¥u¦/§¨9©6ª?«�¬�­=®]¯-°�±K²%³�´�µ�¶¸·-¹�º»�¼�½�¾¿]À�ÁÃÂ�Ä�Å/Æ�Ç/È/ÉÊÌË?Í/ÎÐÏ�Ñ9Ò�Ó�Ô�Õ×Ö�Ø�Ù¸Ú�Û�Ü�Ý�Þ�ß�à/á�â6ã%ä�å�æ6ç%è%é{êVë�ì�í-î3ï)ðòñ�ó3ô-õ�ö

÷ùø�ú�ûýü�þNÿ������������� ���������������� ��!#"�$&% ')(+*�, -�.0/2143�56�798�:+;�<&=�>@?BADC�E F&GIH�JLKNM+O&PRQTS�U+V0WYX�Z�[R\Y]_^9`_a4b0cedfhg�i�j�k0l9monIp�qsr�tvuNw�x&y�zo{&|~}����������

�����~�e� ���I�B���+��� �T���+�0� �����~�� &¡£¢�¤ ¥§¦~¨)©�ª«­¬®B¯�°�±�²T³)´�µ+¶&·�¸º¹h»~¼�½¿¾�À�ÁÃÂYÄ�Å4Æ�Ç0ÈBÉËÊ�Ì)ÍÏÎ�Ð4Ñ�Ò�Ó)Ô�ÕÖ×&Ø&ÙYÚÜÛÞݧß�à&áhâhã�äæåèçêé~ëIì�í�î ïñðóò&ô&õ4ö�÷&ø�ù&úüûþý)ÿ������������ ������������������! #"%$�&('�)

*,+.-0/�1�243�57698;:�<�=�>!?A@�BDC!E�FHGJILK(M.NPORQ�S�T!U�V�WLXZY�[�\�]�^_a`!b�c�d�egf�h�i�jlk.m#n�o(p.qZr,s.t�uwv�x%y.z�{7|!}~����������������7���!���������L���!�����w�����7�����( �¡�¢g£.¤�¥§¦�¨�©«ªL¬®­�¯±°a²a³;´�µ·¶�¸�¹�º±»

¼,½.¾0¿�À�Á4Â�Ã7ÄÆÅ;Ç�È�É�Ê!ËAÌ�ÍÏÎ,Ð�Ñ%ÒÔÓ�Õ ÖØ×�Ù4Ú!Û�Ü�ÝßÞ!àâá�ã%ä�åæ±ç�è�é�ê�ë�ì í!î�ïñðóò�ô%õö�÷#øúù�ûÔü�ýLþ ÿ ��������� ����������������������! #"%$'&)(+*-,/.1032)46587:9<;>=�?-@BADCFEHG%I%J�K�LNMPOQ�RSUTNV�W�XY[Z]\_^a`cbedgfihNjlknm�oNprq�s�tDu6vxwHy�z|{~}N���n�_�����/����a�����U���N�������N�H�F�n�n���r���U�����¡ 3¢l£F¤¦¥)§F¨%©Hª)«�¬l­¯®a°²±!³µ´¶�·�¸�¹Nºn»½¼#¾�¿FÀ�Á6Â

ÃUÄlÅ'Æ�Ç)È6É�ÊÌËÎÍnÏ)Ð�Ñ)Ò�ÓÕÔ�Ö�×�Ø�ÙÛÚcÜ�ݦÞ<ß%àHá'â)ãlä�åçæéè6ê%ë�ì%íFîïñð�ò�ó]ô6õcö�÷�ølù�úNûcüþý ÿ ��������� ������������������� �"!�#�$&%(' )�*+-,/.10325476�839;:�<>=@?5ACB�DFE GIHKJML-N�O3PRQ�S�T3U"V�WYX[Z]\_^�`1a�bced�f3gRh3i3jRk lnmIo�p�q�rFs/tvu3w�x yRz|{�}3~�������� � ��������������_���R�_���3�3�������-�����¡ n¢I£�¤�¥ ¦>§�¨v©�ª¬«&­;®e¯�°3±R²�³3´Rµ�¶�·¸R¹>º�»½¼�¾À¿FÁ1Â�Ã�ÄRÅÇÆ�È3É�Ê�Ë>Ì|ÍÇÎ�ÏÑÐÇÒ�Ó�ÔÖÕ|×nØ]Ù Ú�Û�Ü ÝnÞ�ß

àFá�â�ã�ä�åÖæRçÇè/é�ê�ë]ìRíïî|ð�ñIò�ó-ôöõ�÷�ø�ù5ú�û�ü�ý@þ ÿ������������� ������������������ "!$#&%('*)�+�, -�.0/21�3�4�57698�:<;�=">�?A@CB�D�E�FHGJI�KLNM�O�P�Q�RTS�UWV$XZY\[�]_^a`cbedgfihkjml�npo�q�r�st�uwvyxWz�{}|~���*���J�*�����g�e�����������_����T�*���$�"�J�H���* ¡�¢¤£

¥§¦p¨p©$ªp«c¬®­¯p°�±J²C³�´gµ·¶�¸�¹»ºJ¼Z½Z¾(¿»À$Á�ÂiÃ�Ä*Å�Æ(ÇZȤɻÊ�Ë�ÌJÍ�ÎÐÏHÑmÒyÓÕÔ×ÖÙØ(Ú

ÛJÜ

Page 32: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ÝpÞ�ßJàCá�âgã�ä�å�æ_ç$èêéeë(ì9íHî×ïyð�ñCò*óJô�õ*öø÷�ù¤ú�ûCüý�þCÿ����������� ������� �������

� �������! #"%$'&%( )�*,+.-%/'0�1325476.8:9�;�<>=@?BA.C@D E@FBGIH�J K�L�M'N�O�PRQTS'U'V W�XZY�[3\Z]Z^>_a`bdc�e f�g�h'i�j�kml�n@oqpsr@t�uwv#x y�z�{�|s} ~��Z���w�����'�%�'���������������������%���������� �¡�¢�£�¤a¥¦¨§>©@ª�«�¬a­'®�¯#°5±>²@³µ´·¶ ¸�¹3ºa».¼�½�¾�¿�À�Á� ÃaÄŨÆ>Ç È É�ÊÌË>Í@Î�ÏsÐ'Ñ,Ò�Ó�Ô�Õ�Ö�×�ØaÙÚdÛ�Ü Ý�Þ3ßáà'â�ã#ä%å'æ�ç3èwé·ê ë�ì3íaîaï ð@ñ ò�ó�ô�õ@öq÷sø@ù�úwû�üþý@ÿ����

������� �������������������������� "!$#�%'&�(*)�+�,�-�.�/�0�1324�5"6$798�:3;<>=@?BADC�E3FHG'IJK�L�M"N�O"P�QR"SBTVUXWBY"Z\[�]�^�_"`HaXb�c�dfe�g\hjijkl�m�n"o�p"q�rtsvu�w�x�y{zf|>}$~f�f�@�9��*�����"���H���>���������f����������������� H¡�¢�£¥¤"¦$§©¨«ª�¬9­®*¯�°�±"²�³H´�µ>¶�·¹¸�ºH»�¼�½�¾�¿�À�ÁÂHÃ�Ä�Å¥Æ"Ç�ÈÊÉ�Ë�ÌHÍ>Î�ÏÑÐ�Ò�ÓfÔ�Õ"Ö�×�Ø�Ù�Ú�Û©Ü«Ý�Þ9ßàâá�ã�ä�åæèç�é�ê�ëìHí�î¥ï�ð ñ�ò@óBôõ�öø÷úù�û�ü"ý�þ ÿ��������� ��������� ������������! #"%$�&�'(*),+%-.�/10�2354%68798: ;=<�>�?%@8AB8C DFE�G�H�I JK�L�M�NOQPSRUTWV�XZY\[8]�^`_8acbedSf1g�h�ij�k%l�m no p�q�rsQtSuUvxw�y�z�{ |}�~����Z�5�%�8��8� �F������� ����������S��� ��������� ������� �¡¢�£�¤¦¥%§¦¨ª©�«�¬�­1®�¯°�±�²�³ ´Uµ`¶�·¹¸=º�»`¼¾½À¿ÂÁ�Ã

Ä�Å�Æ�Ç�È�ÉËÊÍÌ�Î*Ï�Ð�Ñ Ò8Ó ÔÖÕ8×�Ø�ÙÚ Û�Ü�ÝÞ`ß�à#áªâ�ã*ä�å�æ çèéÖê`ë8ì�í`î8ï¦ðñ�òôó\õ8ö�÷`ø8ùûúýü þ�ÿ������������ ������������������������� "!$# %'&�(*),+.-�/'0�132.46587:9<;>=@?�ABDC8E'F.G�H�I�J�KML�NDO.P�QDR8S'T.U�V�WYX�Z.[�\�].^�_�`�a.b�c�d.egf,h.i�j�k�lnm.o�pMq�r�st�u�v�w�x�y�z'{}|,~.���'�����������.�M���,�.����� �.�����'���������������� �¡.¢6£�¤�¥�¦�§�¨

©

ª¬«®­.¯�°8±³²µ´�¶*·�¸�¹»º3¼�½¿¾�À>ÁÃÂ:ÄÆÅÈÇÊÉÃËÍÌMÎ8ÏMÐÈÑ8Ò¿Ó'Ô�Õ×Ö'Ø®Ù�ÚµÛ³ÜÞÝÃßáà»â.ã×äÃå�æ}çÃèÞé�ê

ë'ì:íMîÊïñð®òMóõô<ö�÷Þøúù�û×üþý�ÿ������������� ��������������������! #"�$!%'&)(+*-,/.103254!68789;:8<)=!>@?BADCFE�G�HJILKBMON�PRQTS�UV�W'XZY�[]\_^F`ba�c'dReZf�g�hji5kZlLmn8o]p/qsr8tsuwvyxbz|{y}j~/�'���|���������-�3�8�w�y�!���j�����������w���@�L�B���¡ 8¢w£b¤s¥§¦|¨�©sªj«b¬�­!®�¯!°'±�²'³�´�µ@¶�·!¸8¹�º�»j¼_½¿¾�À8ÁOÂ�Ã�ÄyÅ�Æ�Ç|È�É|ÊÌË�ÍsÎ�Ï�Ð|ÑyÒTÓBÔyÕ|Ö�×�Ø�Ù�ÚwÛbÜbÝÞbß|àyájâ/ãjäså3æ�çéè3ê-ë|ì'í�î�ï�ð�ñ_òLóBôöõj÷5ø�ù�ú!û�ü�ý8þ8ÿ���������� ����������������������� ��!�"�#%$'&)(+*,.-0/'1�2 3�4�5�6798;:�<'=�>�?�@ACBEDGF+HEIKJMLNPORQTSMU�V�W)XY�Z\[�]^`_0aEbCc�d%eEfhg�ijlk0m'n�oMp�q�r�st`uwvx�yGz {}|�~;�����M�K�+���h���G���0�l�+�����G���E�����������C�C���¡ G¢+£�¤h¥`¦�§©¨;ª�«E¬+­¯®C°²±'³�´PµE¶�·�¸¹º}»¯¼G½h¾¡¿�ÀÂÁÄÃ�Å�ÆKÇ�È�ÉÊGËÌ�ÍGÎ Ï.Ð�Ñ©Ò©ÓÕÔ'Ö+×hØ}ÙhÚKÛCÜÞÝÄßÕà²áãâ�äæå�çéèhêGëíìïî)ðCñ�òæóïô�õÕö'÷ùøïú�ûÕü�ý�þ ÿ������������� �������������������� �"!$#&%('*),+.-,/10�2�3�465798;:=<&>@?A*B;CEDGF9HGI.JGK�L�MENPO�Q"R�S�TVU&WYX[Z�\�]"^G_a`&b@ced;fhg@ikjml@npo�q"r�sut�vkwyxPz�{m|9}P~������@�"����*�&�����.�@� �&���������.�"��������"�G�a�& @¡m¢@£9¤&¥@¦�§m¨E©uª;«�¬.­*®�¯E°G±9²G³.´Gµ�¶�·�¸¹;º�»½¼¿¾ÁÀ�ÂEÃ�Ä�Å"Æ�Ç�È�É*Ê[ËPÌ9Í�Î�Ï�ÐmÑÒ9Ó½Ô�ÕVÖ�×�ØVÙÛÚ�ÜÞÝ�ß�àuáYâäã�å�æ[ç9è�é�ê�ë�ìÁí*îðïPñ�ò9ó�ôhõ@ö"÷Pø.ùÞú�û;ü@ý9þ�ÿ����������� � ���

���

Page 33: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Adapting Existing Grammars: The XLE Experience

Ronald M. Kaplan and Tracy Holloway King and John T. Maxwell IIIPaloAlto ResearchCenterPaloAlto, CA 94304USA�

kaplan, thking, maxwell � @parc.comAbstract

WereportontheXLE parserandgrammardevelop-mentplatform(Maxwell andKaplan,1993)andde-scribehowabasicLexical FunctionalGrammarforEnglishhasbeenadaptedto two different corpora(newspapertextandcopierrepairtips).

1 Intr oduction

Large-scalegrammardevelopmentplatformsshouldbeableto beusedto developgrammarsfor a widevarietyof purposes.In this paper, we reporton thethe XLE system(Maxwell and Kaplan, 1993), aparserandgrammardevelopmentplatformfor Lex-ical FunctionalGrammars.Wedescribesomeof thestrategiesandnotationaldevicesthatenabletheba-sic English grammardevelopedfor the ParGramproject (Butt et al., 1999; Butt et al., 2002) to beadaptedto two corporawith differentproperties.

1.1 The Corpora

The STANDARD PargramEnglishgrammarcoversthecorephenomenaof English(e.g.,mainandsub-ordinateclauses,noun phrases,adjectivesand ad-verbs,prepositionalphrases,coordination;see(Buttetal.,1999)).Wehavebuilt twodifferentspecializedgrammarson topof this: theEUREKA grammarandtheWSJ grammar.

The EUREKA grammarparsesthe Eurekacor-pusof copierrepairtips, a collectionof documentsoffering suggestionsfor how to diagnoseand fixparticularcopiermalfunctions.Theseinformal andunediteddocumentswerecontributedby copierre-pair technicians,andthecorpusis characterizedbya significantamountof ungrammaticalinput (e.g.,typos,incorrectpunctuation,telegraphicsentences)and much technicalterminology (1). The goal ofparsingthiscorpusis to provideinputto asemanticsand world-knowledgereasoningapplication (Ev-erettet al., 2001).

(1) a. (SOLUTION 2703270) If exhibiting 10-132 faults replacethe pre-fusertransportsensor(Q10-130).

b. (SOLUTION 2724080) 4. Enter into themachinelog, the changesthat have beenmade.

The WSJ grammarcoverstheUPennWall StreetJournal (WSJ) treebanksentences(Marcus et al.,1994). This corpus is characterizedby long sen-tenceswith manydirect quotesandpropernames,(2a). In addition, for evaluationand training pur-poseswealsoparsedaversionof thiscorpusmarkedupwith labeledbracketsandpart-of-speechtags,asin (2b).Riezleretal. (2002)reportonour WSJ pars-ing experiments.

(2) a. But since1981,Kirk HorseInsuranceInc.of Lexington,Ky. hasgrabbeda 20%stakeof themarket.

b. But since 1981, [NP-SBJ Kirk Horse In-suranceInc. of Lexington,Ky.] has/VBZgrabbed/VBN [NPa20%stakeof themar-ket].

2 Priority-based Grammar SpecializationThe XLE systemis designedso that the grammarwriter canbuild specializedgrammarsby both ex-tendingandrestrictinganothergrammar(in ourcasethe basegrammaris the STANDARD Pargram En-glish grammar).An LFG grammaris presentedtothe XLE systemin a priority-orderedsequenceoffiles containingphrase-structurerules, lexical en-tries, abbreviatorymacrosand templates,featuredeclarations,and finite-state transducersfor tok-enizationand morphologicalanalysis.XLE is ap-plied to a single root file holding a CONFIGURA-TION thatidentifiesall theotherfilescontainingrel-evant linguistic specifications,that indicateshow

29

Page 34: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

thosecomponentsareto be assembledinto a com-pletegrammar, andthatspecifiescertainparametersthat controlhow that grammaris to be interpreted.A key ideais that therecanbe only onedefinitionof an item of a given type with a particularname(e.g.,therecanbeonlyoneNPrulealthoughthatsin-gle rulecanhavemanyalternativeexpansions),anditemsin ahigherpriority file overridelowerpriorityitemsof thesametypewith thesamename.Thissetup is similar to thepriority-overrideschemeof theearlierLFG GrammarWriter’s Workbench(KaplanandMaxwell, 1996).

Thisarrangementmakesit relativelyeasyto con-struct a specializedgrammarfrom a pre-existingstandard.The specializedgrammaris definedbya CONFIGURATION in its own root file that speci-fies the relevantSTANDARD grammarfiles aswellas the new files for the specializedgrammar. Thefiles for the specializedgrammarcanalso containitemsof differenttypes(phrase-structurerules,lex-ical entries,templates,etc.), and they are orderedwith higherpriority thantheSTANDARD files.

Considertheconfigurationfor theEUREKA gram-mar. It specifiesall of theSTANDARD grammarfilesaswell asits own rule, template,lexicon,andmor-phologyfiles. A partof this configurationis shownin (3) (thenotationtemplates.lfgaresharedby all thelanguages’grammars,not just English).

(3) FILES ../standard/english-lexicons.lfg../standard/english-rules.lfg../standard/english-templates.lfg../../common/notationtemplates.lfgenglish-eureka-morphconfigeureka-lexicons.lfgeureka-rules.lfgeureka-templates.lfg

This configurationspecifiesthat theEUREKA rules,templates,and lexical entries are given priorityover the STANDARD items by putting the spe-cial EUREKA files at the end of the list. Thus, ifthe../standard/english-rules.lfgandeureka-rules.lfgfilesbothcontaina ruleexpandingtheNPcategory,theonefromtheSTANDARD file will bediscardedinfavor of theEUREKA rule.

In the following subsections,we provideseveralillustrationsof howsimpleoverridinghasbeenusedfor theEUREKA andWSJ grammarextensions.

2.1 RulesThe overrideconventionmakesit possibleto: addrules(e.g.,for new or idiosyncraticconstructions);deleterules(e.g.,toblockconstructionsnotfoundinthenewcorpus);andmodify rulesto allow differentdaughtersequences.

Rulesmayneedto beaddedto allow for corpus-specificconstructions.This is illustratedin the EU-REKA corpusby theidentifier informationthatpre-cedeseachsentence,asin (1). In orderto parsethissubstring,anewcategory(FIELD) wasdefinedwithan expansionthat coversthe identifier informationfollowed by theusualROOT categoryof the STAN-DARD grammar. The top-level categoryis one oftheparametersof a configuration,andthe EUREKA

CONFIGURATION specifiesthatFIELD insteadof theSTANDARD ROOT is the start-symbolof the gram-mar. Thusthe EUREKA grammarproducesthe treein (4) andfunctional-structure in (5) for (1a).

(4) FIELD

LP EURHEAD ID SUB-ID RP ROOT

( SOLUTION 27032 70 ) � � �(5) �������������

PRED � replace� SUBJ, OBJ���SUBJ [ � � � ]OBJ [ � � � ]� � �FIELD solutionTIP-ID 27032SUB-TIP-ID 70

� �����������!It is unusualin practiceto needto deletea rule,

i.e., to eliminatecompletelythe possibility of ex-pandinga given categoryof the STANDARD gram-mar. This is generallyonly motivatedwhenthespe-cializedgrammarappliesto adomainwherecertainconstructionsare rarely encountered,if at all. Al-thoughtherehasbeennoneedto deleterulesfor theEUREKA andWSJ corpora,theoverrideconventionalsoprovidesanaturalwayof achievingthiseffect.For example,topicalizationis extremelyrarein thetheEurekacorpusandtheSTANDARD topicalizationrulesometimesintroducesparsinginefficiency. Thiscanbeavoidedby havingthehighpriority EUREKA

file replacetheSTANDARD rulewith theonein (6).

(6) CPtop "$# .

ThisvacuousruleexpandstheCPtopcategoryto theemptylanguage,thelanguagecontainingnostrings;

30

Page 35: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

so, this categoryis effectively removedfrom thegrammar.

Perhapsthe most common changeis to makemodificationsto thebehaviorof existingrules.Themostdirect way of doing this is simply to defineanew, higherpriority expansionof thesameleft-handcategory. SinceXLE only allowsa singlerule for agivencategory, theold rule is discardedandthenewonecomesinto play. The new rule canbe arbitrar-ily differentfrom theSTANDARD one,butthis is nottypically thecase.It is muchmorecommonthatthespecializedversionincorporatesmostof thebehav-ior of theoriginal,with minor extensionsor restric-tions.Oneway of producingthemodifiedbehavioris to createa newrule that includesa copyof someor all of theSTANDARD rule’s right sidealongwithnewmaterial,andto give thenewdefinitionhigherpriority thantheold.Forexample,pluralsin theEu-rekacorpuscanbe formedby theadditionof ’s in-steadof theusuals, asin (7).

(7) (CAUSE2741610)A 7mfd invertermotorca-pacitorwasinstalledonanunknownnumberofUDH’s.

In orderto allow for this,theN rulewasrewrittentoallow a PL markerto optionallyoccurafteranyN,asin (8).

(8) N "%# copyof STANDARD N rule(PL)

As a resultof this rule modification,UDH’s in (7)will havethetreeandfunctional-structure in (9).

(9) a. N

PART PL

UDH ’s

b. &PRED � UDH �NUM pl '

Copyingmaterialfrom oneversionto anotherisperhapsreasonablefor relativelystableandsimplerules,like theN rule,butthiscancausemaintainabil-ity problemswith complicatedrules in the STAN-DARD grammarthatareupdatedfrequently. An al-ternativestrategyis to movethebodyof theSTAN-DARD N rule to a differentrule,e.g.,Nbody, whichin turnis calledby theN rulein boththeSTANDARD

andEUREKA grammars.TheNbodycategorycanbesupressedin thetreestructureby invoking this ruleasa macro(notationallyindicatedas@Nbody).

(10) N "%# @Nbody (PL).

Often the necessarymodification can be madesimply by redefininga macrothatexistingrulesal-readyinvoke.ConsidertheROOT rule, in (11).

(11)ROOT "%# ( @DECL-BODY @DECL-PUNCT)@INT-BODY @INT-PUNCT)@HEADER * .

In the STANDARD grammar, the DECL-PUNCT

macro is definedas in (12a). However, this mustbe modified in the EUREKA grammarbecausethepunctuationis much sloppier and often doesnotoccurat all; theEUREKA versionis shownin (12b).

(12) a. DECL-PUNCT = ( PERIOD)EXCL-POINT * .

b. DECL-PUNCT = ( ( PERIOD)EXCL-POINT)COLON)SEMI-COLON * ).

The modular specificationsthat macrosand tem-platesprovide allow rule behaviorto be modifiedwithout havingto copythepartsof therule thatdonot change.

XLE also has a mechanism for systemati-cally modifying the behavior of all rules: theMETARULEMACRO. For example, in order toparselabeledbracketedinput, as in (2b), the WSJ

grammar was altered so that constituentscouldoptionally be surrounded by the appropriatelylabeledbrackets.TheMETARULEMACRO is appliedto eachrule in thegrammarandproducesasoutputa modified version of that rule. This is used inthe STANDARD grammarfor coordinationand toallow quotemarksto surroundanyconstituent.TheMETARULEMACRO is redefinedfor the WSJ to addthelabeledbracketingpossibilitiesfor eachrule,asshownin (13).

(13) METARULEMACRO( CAT BASECAT RHS) =( LSB LABEL[ BASECAT] CAT RSB)copyof STANDARD coordination)copyof STANDARD surroundingquote * .

The CAT, BASECAT, and RHS are argumentstothe METARULEMACRO that are instantiatedto dif-ferentvaluesfor eachrule. RHS is instantiatedtothe right-handsideof the rule, i.e., the rule expan-sion. CAT and BASECAT are two ways of repre-sentingtheleft-handsideof therule.Forsimplecat-egoriesthe CAT and BASECAT arethe same(e.g.

31

Page 36: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

NP for theNP rule). XLE alsoallows for complexcategorysymbolsto specializetheexpansionof par-ticular categoriesin particularcontexts.For exam-ple, theVP rule is parameterizedfor theform of itscomplementandits own form, so thatVP[perf,fin]is one of the complex VP categories.When theMETARULEMACRO appliesto rules with complexleft-sidecategories,CAT refersto thecategoryin-cluding theparametersandthe BASECAT referstothecategorywithout theparameters.For theVP ex-ample, CAT is VP[perf,fin] and BASECAT is VP.

In the definition in (13), LSB and RSB parsethebracketsthemselves,while the LABEL[ BASECAT]parsesthe label in thebracketingandmatchesit tothelabelin thetree(NPin (2b));theconsituentitselfis the CAT. Thus,a label-bracketedNP is assignedthestructurein (14).

(14) NP

LSB LABEL[NP] NP RSB

[ NP-SBJ Kirk Horse � � � ]

Theseexamplesillustratehow theprioritizedre-definitionof rulesandmacroshasenabledusto in-corporatetheSTANDARD rulesin grammarsthataretunedto the specialpropertiesof the EUREKA andWSJ corpora.

2.2 Lexical Entries

Justasfor rules,XLE’s overrideconventionsmakeit possibleto: addnewlexical itemsor newpart-of-speechsubentriesfor existing lexical items;deletelexical items;andmodify lexical items.In additionto the basic priority overrides,XLE providesfor“edit lexical entries” (KaplanandNewman,1997)that give finer control over the constructionof thelexicon.Editentrieswereintroducedasawayof rec-oncilinginformationfrom lexicaldatabasesof vary-ing degreesof quality, but they arealsohelpful intailoring a STANDARD lexiconto a specializedcor-pus.Whenworking on specializedcorpora,suchastheEurekacorpus,modificationsto the lexiconareextremelyimportantfor correctlyhandlingtechni-calterminologyandeliminatingwordsensesthatarenotappropriatefor thedomain.

Higher-priority editlexicalentriesprovidefor op-eratorsthat modify the definitionsfound in lower-priority entries.The operatorscan:adda subentry(+); deletea subentry( " ); replacea subentry(!);or retainexisting subentries(=). For example,the

STANDARD grammarmighthaveanentryfor buttonasin (15).

(15) button !V @(V-SUBJ-OBJ %stem);!N @(NOUN %stem);ETC.

However, theEUREKA grammarmightnotneedtheV entrybut might requirea specialpartnameN en-try. Assumingthat the EUREKA lexiconsaregivenpriority over the STANDARD lexicons,the entry in(16)wouldaccomplishthis.

(16) button " V ;+N @(PARTNAME %stem);ETC.

Notethatthelexicalentriesin (15)and(16)endwithETC. This is alsopart of the edit lexical entry sys-tem.It indicatesthatotherlower-priority definitionsof that lexical item will be retainedin addition tothenewentries.Forexample,if in anotherEUREKA

lexicontherewasanadjectiveentryfor buttonwithETC, theV, N, andA entrieswouldall beused.Thealternativeto ETC is ONLY whichindicatesthatonlythenewentryis to beused.In ourbuttonexample,ifanadjectiveentrywasaddedwith ONLY, theV andN entrieswould beremoved,assumingthat thead-jectiveentryoccurredin thehighestpriority lexicon.This machineryprovidesa powerful tool for build-ing specializedlexiconswithout havingto alter theSTANDARD lexicons.

The EUREKA corpuscontainsa large numberofnamesof copierparts.Due to their particularsyn-taxandto post-syntacticprocessingrequirements,aspeciallexical entryis addedfor eachpartname.Inaddition,the regularnounparseof theseentriesisdeletedbecausewhenevertheyoccurin thecorpusthey arepart names.A samplelexical is shownin(17); the‘ is theescapecharacterfor thespace.

(17) separator‘finger!PART-NAME @(PART-NAME %stem);" N;ETC.

Thefirst line in (17)statesthatseparatorfingercanbea PART NAME andwhenit is, it callsa templatePART-NAME that providesrelevantinformationforthe functional-structure. The secondline removesthe N entry, if any, assignalledby the " beforethecategoryname.

32

Page 37: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Becauseof thenon-contextfreenatureof LexicalFunctionalGrammar, it sometimeshappensthatex-tensionsin onepart of the grammarrequirea cor-respondingadjustmentin otherrulesor lexical en-tries. Consideragain the EUREKA ’s plurals. Thepart-nameUDH is singularwhen it appearswith-out the ’s andthusthemorphologicaltag+Sgis ap-pendedto it. In theSTANDARD grammar, thetag+Sghasalexicalentryasin (18a)whichstatesthat+Sgisof categoryNNUM andassignssgto its NUM. How-ever, if this is usedin theEUREKA grammar, thesgNUM specificationwill clashwith thepl NUM spec-ification whenUDH appearswith ’s, asseenin (7).Thus,a new entry for +Sg is neededwhich hassgasadefaultvalue,asin (18b).Thefirst line of (18b)statesthat NUM must exist but doesnot specify avalue,while thesecondline optionallysuppliesasgvalueto NUM; whenthe ’s is used,this optiondoesnotapplysincetheform alreadyhasapl NUM value.

(18) a. +Sg NNUM ( + NUM)=sg

b. +Sg NNUM ( + NUM)(( + NUM)=sg)

3 Tokenizing and Morphological AnalysisTokenizationand morphologicalanalysisin XLEarecarriedoutbymeansof finitestatetransductions.The STANDARD tokenizingtransducerencodesthepunctuationconventionsof normal English text,which is adequatefor manyapplications.However,theEurekaandWSJcorporaincludestringsthatmustbetokenizedin non-standardways.TheEurekapartidentifiershaveinternalpunctuationthatwouldnor-mally causeastringtobebrokenup(e.g.thehyphenin PL1-B7),andthe WSJ corpusis markedup withlabeledbracketsandpart-of-speechtagsthat mustalso receivespecialtreatment.An exampleof theWSJ mark-upis seenin (19).

(19) [NP-SBJLloyd’s, oncea pillar of the worldinsurance market,] is/VBZ being/VBGshaken/VBN to its very foundation.

Part-of-speechtagsappearin a distinctive format,beginningwith a / andendingwith a , with thein-terveningmaterialindicatingthecontentof the tag(VBZfor finite 3rdsingularverb,VBGfor aprogres-sive,VBN for apassive,etc.).Thetokenizingtrans-ducermustrecognizethis patternandsplit the tagsoff asseparatetokens.Thetag-tokensmustbeavail-able to filter the outputof the morphologicalana-lyzer sothatonly verbalformsarecompatiblewith

thetagsin thisexampleandtheadjectivalreadingofshakenis thereforeblocked.

XLE tokenizing transducersare compiled fromspecificationsexpressedin thesophisticatedXeroxfinite statecalculus(BeesleyandKarttunen,2002).The Xerox calculusincludesthe composition,ig-nore,andsubstitutionoperatordiscussedby KaplanandKay (1994)andthe priority-union operatorofKaplanandNewman(1997).The specializedtok-enizersareconstructedby usingtheseoperatorstocombinethe STANDARD specificationwith expres-sionsthat extendor restrict the standardbehavior.Forexample,theignoreoperatoris appliedto allowthepart-of-speechinformationto bepassedthroughto themorphologywithout interruptingthestandardpatternsof Englishpunctuation.

XLE also allows separatelycompiledtransduc-ers to be combinedat run-time by the operationsof priority-union,composition,andunion.Priority-unionwasusedtosupplementthestandardmorphol-ogywith specialized“guessing”transducersthatap-ply only to tokensthat would otherwisebe unrec-ognized.Thus,a finite-stateguesserwas addedtoidentify Eurekafault numbers(09-425),adjustmentnumbers(12-23),partnumbers(606K2100),partlistnumbers(PL1-B7), repairnumbers(2.4), tagnum-bers(P-102),anddiagnosticcodenumbers(dC131).Compositionwasusedto apply the part-of-speechfiltering transducerto theoutputof themorpholog-ical analyzer, and union providedan easyway ofaddingnew, corpus-specificterminology.

4 Optimality MarksXLE supportsa versionof Optimality Theory(OT)(PrinceandSmolensky, 1993)which is usedto rankananalysisrelativetootherpossibleanalyses(Franketal.,2001).In general,thisis usedwithin aspecificgrammarto preferor dispreferaconstruction.How-ever, it canalsobe usedin grammarextensionstodeleteor includerulesor partsof rules.

The XLE implementationof OT works as fol-lows.1 OT marksareplacedin thegrammarandareassociatedwith particular rules, parts of rules, orlexical entries.Thesemarksarethenrankedin thegrammarCONFIGURATION. In additionto a simplerankingof constraintswhich statesthata construc-tion with a given OT mark is (dis)preferedto one

1The actualXLE OT implementationis morecomplicatedthan this, allowing for UNGRAMMATICAL and STOPPOINT

marksas well. Only OT marksthat are associatedwith NO-GOOD areof interesthere.For a full description,see(Franketal., 2001).

33

Page 38: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

without it, XLE allowsthemarksto bespecifiedasNOGOOD. A rule or rule disjunctwhich hasa NO-GOOD OT mark associatedwith it will be ignoredby XLE. This canbe usedfor grammarextensionsin thatit allowsastandardgrammarto anticipatethevariationsrequiredby specialcorporawithoutusingthemin normalcircumstances.

Considerthe exampleof the EUREKA ’s pluralsdiscussedin section2.1. Insteadof rewriting theNrule in the EUREKA grammar, it would bepossibleto modify it in theSTANDARD grammarandincludeanOT mark,asin (20).

(20) N "%# original STANDARD N rules(PL: @(OT-MARK EUR-PLURAL)).

The CONFIGURATION files of the STANDARD andEUREKA grammarswould differ in that the STAN-DARD grammarwould rank the EUR-PLURAL OTmark as NOGOOD, as in (21a),while the EUREKA

grammarwould simply not rank the mark, as in(21b).

(21) a. STANDARD optimality order:EUR-PLURAL NOGOOD � � �

b. EUREKA optimalityorder:NOGOOD � � �

GiventheOT marks,it wouldbepossibleto haveone large grammarthat is specializedby differentOT rankingsto producethe STANDARD, EUREKA,andWSJ variants.However, from a grammarwrit-ing perspectivethis is not a desirablesolutionbe-causeit becomesdifficult to keep track of whichconstructionsbelong to standardEnglish and aresharedamongall thespecializationsandwhich arecorpus-specific.In addition,it doesnotdistinguishacoresetof slowly changinglinguisticspecificationsfor thebasicpatternsof thelanguage,andthusdoesnot providea stablefoundationthat the writers ofmorespecializedgrammarscanrely on.

5 Maintenancewith Grammar Extensions

Maintenanceis a seriousissuefor any large-scalegrammardevelopmentactivity, andthemaintenanceproblemsarecompoundedwhenmultiple versionsarebeingcreatedperhapsby severaldifferentgram-marwriters.Our STANDARD grammaris now quitematureandcoversall the linguistically significantconstructionsandmostotherconstructionsthatwe

haveencounteredin previouscorpusanalysis.How-ever, everynowandthen,anewcorpus,evenaspe-cializedone,will evidencea standardconstructionthat hasnot previouslybeenaccountedfor. If spe-cializedgrammarswerewritten by copyingall theSTANDARD files andthenmodifying them,the im-plementationof new standardconstructionswouldtendto appearonly in thespecializedgrammar. Ourtechniquesfor minimizing the amountof copyingencouragesusto implementnewconstructionsin theSTANDARD grammarandthismakesthemavailableto all otherspecializations.

If a newversionof a rule for a specializedgram-maris createdby copyingthecorrespondingSTAN-DARD rule, changeslater madeto the specialrulewill notautomaticallybereflectedin theSTANDARD

grammar, andvice versa.This is thedesiredbehav-ior whenaddingunusual,corpus-specificconstruc-tions. However, if the non-corpusspecificpartsofthenewrule aremodified,thesemodificationswillnot migrateto the STANDARD grammar. To avoidthis problem,the smallestrule possibleshouldbemodifiedin the specializedgrammar, e.g.,modify-ing the N headrule insteadof the entire NP. Forthisreason,havinghighly modularizedrulesandus-ing macrosandtemplateshelpsin grammarmainte-nancebothwithin agrammarandacrossspecializedgrammarextensions.

As seenabove,the XLE grammardevelopmentplatformprovidesanumberof mechanismsto allowfor grammarextensionswithout altering the core(STANDARD) grammar. However, therearestill ar-easthat could use improvement.For example,asmentionedin section2, the CONFIGURATION filestateswhich other files the grammarincludesandhowtheyareprioritized.TheCONFIGURATION con-tainsother informationsuchasdeclarationsof thegovernablegrammaticalfunctions,the distributivefeatures,etc. As this information rarely changeswith grammarextensions,it would be helpful foran extensionconfigurationto incorporateby refer-encesuchadditionalparametersof the STANDARD

configuration.Currentlythesedeclarationsmustbecopiedinto eachCONFIGURATION.

6 Discussionand Conclusion

As a resultof the strategiesandnotationaldevicesoutlined above, our specializedgrammarssharesubstantialportionsof the pre-existingSTANDARD

grammar. Thestatisticsin table(22)giveanindica-tion of the sizeof the STANDARD grammarandof

34

Page 39: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

theadditionalmaterialrequiredfor theEUREKA andWSJ specializations.As canbeseenfrom this table,thespecializedgrammarsrequirea relativelysmallnumberof rulescomparedto therulesin theSTAN-DARD grammar. Thenumberof lines that the rulesandlexicalentriestakeupalsoprovidesameasureoftherelativesizeof thespecifications.TheWSJ lexi-consincludemanytitles andpropernounsthatmayultimatelybemovedto theSTANDARD files.Theta-blealsoshowsthenumberof filescalledbytheCON-FIGURATION, asanotherindicationof thesizeof thespecifications.Thisnumberissomewhatarbitraryasseparatefiles canbe combinedinto a singlemulti-sectionedfile, althoughthisis likely to reducemain-tainabilityandreadability.

(22)

STANDARD EUREKA WSJ

rules 310 32 14lines:

rules 6,539 425 894lexicons 44,879 5,565 15,135

files 14 5 8

Thegrammarscompileinto acollectionof finite-statemachineswith the numberof statesandarcslistedin table(23).TheWSJ grammarcompilesintothelargestdatastructures,mainlybecauseof itsabil-ity to parselabeledbracketedstringsand part-of-speechtags,(2b). This sizeincreaseis theresultofaddingonedisjunct in the METARULEMACRO andhencereflectsonly aminor grammarchange.

(23)STANDARD EUREKA WSJ

states 4,935 5,132 8,759arcs 13,268 13,639 19,695

In sum,the grammarspecializationsystemusedin XLE hasbeenquitesucessfulin developingcor-pusspecificgrammarsusingtheSTANDARD Englishgrammarasabasis.A significantbenefitcomesfrombeingableto distinguishtruly unusualconstructionsthatexistonly in thespecializedgrammarfromthosethatare(or shouldbe) in the STANDARD grammar.Thisallowsidiosyncraticinformationto remainin aspecializedgrammarwhile all thespecializedgram-marsbenefitfrom andcontributeto the continuingdevelopmentof theSTANDARD grammar.

ReferencesK. Beesleyand L. Karttunen. 2002. Finite-State

Morphology:Xerox ToolsandTechniques. Cam-bridgeUniversityPress.To Appear.

M. Butt, T.H. King, M.-E. Nino, and F. Segond.1999. A Grammar Writer’s Cookbook. CSLIPublications,Stanford,CA.

M. Butt, H. Dyvik, T.H. King, H. Masuichi, andC. Rohrer. 2002. The parallelgrammarproject.In Proceedingsof COLING 2002. WorkshoponGrammarEngineeringandEvaluation.

J. Everett, D. Bobrow, R. Stolle, R. Crouch,V. de Paiva,C. Condoravdi,M. van den Berg,andL. Polanyi. 2001. Making ontologiesworkfor resolving redundanciesacross documents.Communicationsof theACM, 45:55–60.

A. Frank,T. H. King, J.Kuhn,andJ.T. Maxwell III.2001. Optimality theory style constraintrank-ing in large-scaleLFG grammars.In PeterSells,editor, Formal andEmpirical Issuesin Optimal-ity TheoreticSyntax. CSLIPublications,Stanford,CA.

R. KaplanandM. Kay. 1994. Regularmodelsofphonologicalrule systems.ComputationalLin-guistics, 20:331–378.

R. Kaplan and J. Maxwell. 1996. LFG Gram-marWriter’sWorkbench.Systemdocumentationmanual;availableon-lineatPARC.

R. KaplanandP. Newman.1997. Lexical resourceconciliationin theXeroxLinguisticEnvironment.In Proceedingsof the ACL Workshopon Com-putationalEnvironmentsfor GrammarDevelop-mentandEngineering.

M. Marcus,G. Kim, M. A. Marcinkiewicz,R. Mac-Intyre, A. Bies, M. Ferguson, K. Katz, andB. Schasberger. 1994. The Penntreebank:An-notativepredicateargumentstructure. In ARPAHumanLanguageTechnologyWorkshop.

J.Maxwell andR. Kaplan. 1993.Theinterfacebe-tweenphrasalandfunctionalconstraints.Compu-tational Lingusitics, 19:571–589.

A. PrinceandP. Smolensky. 1993. Optimality the-ory: Constraintinteractionin generativegram-mar. RuCCSTechnicalReport#2, RutgersUni-versity.

S. Riezler, T.H. King, R. Kaplan,D. Crouch,J. T.Maxwell, III, and M. Johnson. 2002. ParsingtheWall StreetJournalusinga lexical-functionalgrammar and discriminative estimation tech-niques.In Proceedingsof theAnnualMeetingofthe Associationfor ComputationalLinguistics,Universityof Pennsylvania.

35

Page 40: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Coping with problemsin grammars automatically extracted fr om tr eebanks

Carlos A. ProloComputerandInformationScienceDepartment

Universityof PennsylvaniaSuite400A,3401WalnutStreet

Philadelphia,PA, USA, [email protected]

AbstractWe report in this paperon an experimenton auto-maticextractionof aTreeAdjoining Grammarfromthe WSJcorpusof the PennTreebank.We useanautomatictool developedby (Xia, 2001) properlyadaptedto ourparticularneed.Ratherthanaddress-ing generalaspectsof the automaticextractionwefocus on the problemswe have found to extract alinguistically(andcomputationally)soundgrammarandapproachesto handlethem.

1 Intr oductionMuch linguistic researchis oriented to findinggeneralprinciples for natural language,classify-ing linguistic phenomena,building regular mod-els (e.g.,grammars)for thewell-behaved (or well-understood)partof languagesandstudyingremain-ing “interesting” problemsin a compartmentalizedway. With theavailability of largenaturallanguagecorporaannotatedfor syntacticstructure,the tree-banks,e.g.,(Marcuset al., 1993),automaticgram-mar extraction becamepossible(Chenand Vijay-Shanker, 2000; Xia, 1999). Suddenly, grammarsstarted being extracted with an attempt to have“full” coverageof theconstructionsin acertainlan-guage(of course,to theextentthattheusedcorporarepresentsthelanguage)andthatimmediatelyposesa question:If we do not knowhow to modelmanyphenomenagrammaticallyhowcanthat bethat weare extractingsuch a wide-coverage grammar?.

To answerthat questionwe have to starta newthreadat theedgeof linguisticsandcomputationallinguistics.Morethannumbersto expresscoverage,we have to startanalyzingthe quality of automat-ically generatedgrammars,identifying extractionproblemsanduncoveringwhateversolutionsarebe-ing givenfor them,however interestingor ugly theymight be,challengingthecurrentparadigmsof lin-guisticresearchto provideanswersfor theproblemson a “by-need”basis.

In this paperwe report on a particular experi-enceof automaticextractionof anEnglishgrammarfrom the WSJcorpusof the PennTreebank(PTB)(Marcuset al., 1994)1 usingTreeAdjoining Gram-mar(TAGs,(JoshiandSchabes,1997)).We useanautomatictool developedby (Xia, 2001) properlyadaptedto our particularneedsandfocuson someproblemswe have found to extract a linguistically(andcomputationally)soundgrammarand the so-lutions we gave to them. The list of problemsisa sample,far from beingexhaustive2 Likewise, thesolutionswill notalwaysbesatisfactory.

In Section2 weintroducethemethodof grammarextractionemployed.TheproblemsarediscussedinSection3. Weconcludein Section4.

2 The extracted grammar

2.1 TAGs

A TAG is asetof lexicalizedelementarytr eesthatcanbecombined,throughtheoperationsof tr eead-junction and tr ee substitution, to derive syntac-tic structuresfor sentences.We follow a commonapproachto grammardevelopmentfor naturallan-guageusing TAGs, underwhich, driven by local-ity principles,eachelementarytreefor a given lex-ical headis expectedto containits projection,andslotsfor its arguments(e.g.,(Frank,2002)). Figure1 shows typical grammartemplatetreesthatcanbeselectedby lexical itemsandcombinedto generatethestructurein Figure2. Thederivationtree, to theright, containsthe history of the treegrafting pro-cessthatgeneratedthederivedtree, to theleft.3

1Weassumesomefamiliarity with thebasicnotationsin thePTB asin (Marcuset al., 1994).

2(Prolo,2002)includesamorecomprehensive anddetaileddiscussionof grammarextractionalternativesandproblems.

3For amorecomprehensive introductionto TAGsandLexi-calizedTAGswereferthereaderto (JoshiandSchabes,1997).

36

Page 41: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

NP

N

S

NP

V NP

VP

VP

VPPP

P NP

DT

NP

* NP*

np vt detppright

Figure1: An exampleof TreeAdjoining Grammar

DT

NP

NP*

N

PP

P

*VP

V NP

N

NP

N

VP

S

[John]

[saw]

[Mary]

[from]

[the]

[window]

Derived tree Derivation tree

vt[saw]

np[John] np[Mary] pp[from]

np[window]

det[the]

Figure2: Derivationof JohnsawMary fromthewindow

2.2 LexTract

Given an annotatedsentencefrom the PTB as in-put Xia’s LexTracttool (Xia, 1999;Xia, 2001)firstexecutesa rebracketing. More precisely, additionalnodesareinsertedto separateargumentsandmod-ifiers andto structurethe modifying processasbi-nary branching.A typical rebracketedPTB treeisshown in Figure3,4 in whichwehavedistinguishedthetreenodesinsertedby LexTract.

Thesecondstageis theextractionof thegrammartreespropershown in Figure 4. In particular, re-cursive modifier structureshave to be detectedandfactoredoutof thederivedtreeto composetheaux-iliary trees,the restbecomingan initial tree. Theprocessis recursive also in the sensethat factoredsubtreestructuresstill undergo thespinningoff pro-cessuntil wehaveall modifierswith theirown trees,all theargumentsof a headassubstitutionnodesofthe treecontainingtheir head,andthematerialun-der the argumentnodesdefining additional initialtreesfor themselves. Auxiliary treesareextractedfrom parent-childpairswith matchinglabelsif thechild is electedtheparent’s headandthechild’s sib-ling is marked as modifier: the parentis mappedintoarootof anauxiliarytree,thehead-childinto its

4Figures3 and4 arethanksto FeiXia. Wearealsogratefulto her for allowing usto useLexTractandmake changesto itssourcecodeto customizeto ourneeds.

foot,with thesiblingsubtree(afterbeingrecursivelyprocessed)beingcarriedtogetherinto theauxiliarytree.Noticethattheauxiliary treesarethereforeei-therstrictly right or left branching,the foot alwaysimmediatelyunderthe root node. Other kinds ofauxiliary treesarethereforenotallowed.

VBP

draft

NP

NNS

policies

VBG

using

εNP

NNS

pens

VP

S

IN

at

NP

NNP

FNX

PRP

theyRB

still

PP−LOC

NP−SBJ

ADVP

S−MNR

VPNP−SBJ

VP

S

VP

Figure3: LexTractrebracketingstage

To extract a grammarwith Xia’s tool onehastodefinetablesfor finding: the headchild of a con-stituentexpansion;which of the siblingsof a headareacceptablearguments;andwhichconstituentla-belsareplausiblemodifiersof another. Specialpro-visionsaremadefor handlingcoordination.For ad-ditional informationsee(Xia, 2001). In this paperwereferto (Xia, 1999)’s tablesettingsandextractedgrammar, which we usedas our startingpoint, asXia’ssample. Weusedacustomizedversionof Lex-Tract,plusadditionalpre-processingof thePTB in-putandpost-processingof theextractedtrees.

3 Extraction ProblemsExtractionproblemsarisefrom severalsources,in-cluding: (1) lack of properlinguistic account,5 (2)the(PennTreebank)annotationstyle, (3) the(Lex-Tract) extraction tool, (4) possibleunsuitabilityofthe (TAG) model, and (5) annotationerrors. Werefrainedfrom makinga rigid classificationof theproblemswe presentaccordingto thesesources.Inparticularit is often difficult to decidewhethertoblamesources(1), (3), or (5) for a certainproblem.We will not discussin this paperproblemsduetoannotationerrors. As for the PTB style problemswe only discussone,thefirst listedbelow.

5Here included the (occasional)inability on the part ofgrammardevelopersto find or makeuseof anexistingaccount.

37

Page 42: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

#7#5

#6

PP

IN

at

S

still

S2.t

RB

ADVP

draft

VBP

NNS

NP3.b

VP3.t

VP2.t

VBG

S3

VP

NP1.t

FNX

NNP

NP1.b NP2.b

PRP

VP3.b

NP5.t

NP5.b

VP1.b

VP1.t

S2.b

1.b

NP2.t

NP3.t

VP2.b

NP4

ε

5

#5

#2#3

#4

#1

S1.t

#5

they

policiesusing NNS

pens

#8

#5:

policies

NNS S

VP

VP*

NP

ε VBG

using

VP

NP

#7:

#1:S

S*PP

IN NP

at

NP

NNP

FNX

#2:

#6: #8:

NP

NNS

pens

NP

S

VP

NPVBP

draft

NP

NP

PRP

#3:

they

#4:

ADVP VP*

RB

still

VP

a) Input treedecomposition b) Extractedelementarytrees

Figure4: LexTractextractionstage

(S-3 (NP-SBJ (PRP We))(VP (VBP make)

(SBAR-NOM (WHNP-1 (WP what))(S we know

how to make))))

a)As asententialclausein thePTB

(S-3 (NP-SBJ (PRP We))(VP (VBP make)

(NP (NP (WP what))(SBAR (WHNP-1 (-NONE- 0))

(S we know ...)))))b) As aNounphraseafterpre-processed

Figure5: Freerelativesin theTreebank

3.1 FreeRelatives

Freerelatives are annotatedin the PennTreebankas sententialcomplementsas in Figure 5.a. Theextractedtree correspondingto the occurrenceof“make” would be of a verb that takes a sententialcomplement(SBAR). Thisdoesnotseemto becor-rect6, asthepropersubcategorizationof theverboc-currenceis transitive.

In fact, freerelativesmayoccurwherever anNPargumentmay occur. So, the only reasonableex-traction accountconsistentwith maintainingthemas SBARs would be one in which every NP sub-stitution nodein an extractedtreewould admit the

6In bothstandardaccountsfor freerelatives,theHeadAc-count(e.g.,(BresnanandGrimshaw, 1978))andtheCompAc-count(e.g.,(Groosandvon Riemsdijk,1979)),commonlydis-cussedin theliterature,thepresenceof theNP(or DP) is clear.

existenceof a counterparttree,identicalto thefirst,exceptthat theNP argumentlabel is replacedwithan SBAR. Insteadwe optedto reflecttheNP char-acterof thefreerelativesby pre-processingthecor-pus (using the Head-analysis, for practicalconve-nience). The annotatedexample is then automat-ically replacedwith the one in Figure 5.b. Othercasesof free-relatives (non-NP) are rare and notlikely to interferewith verbsubcategorization.

3.2 Wh percolation upIn thePennTreebankthesameconstituentis anno-tatedwith differentsyntacticcategoriesdependingon whetherit possessesor not the wh feature. Forinstance,a regular noun phrasehas the syntacticcategory NP, whereaswhen the constituentis wh-marked,andis in thelandingsiteof wh-movement,it carriesthe label WHNP.7 While that might lookappealingsincethe two constituentsseemto havedistinct distributional properties,it posesa designproblem. While regular constituentsinherit theirsyntacticcategorial feature(i.e. their label) fromtheir heads,wh projectionsareoften formedby in-heritancefrom their modifiers. For instance:“thefather” is an NP, but modifiedby a wh expression(“the fatherof whom”, “whosefather”, “which fa-ther”), it becomesa WHNP. The only solutionweseeis to allow for nounsandNPsto freelyprojectupto WHNPsduringextraction.8 Ontheotherhand,in

7Whentheconstituentis not wh-moved,it is correctlypre-servedasanNP, as“what” in “Who atewhat?”.

8Of courseanothersimplesolutionwould be merging thewh constituentswith their non-whcounterparts.

38

Page 43: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(NP (UCP (NN construction)(CC and)(JJ commercial))

(NNS loans))

a) NP modifiers

(VP (VB be)(UCP-PRD (NP (CD 35))

(CC or)(ADJP (JJR older))))

b) non-verbal Predicates

(VP (VB take)(NP (NN effect))(UCP-TMP (ADVP 96 days later)

(, ,)(CC or)(PP in early February)))

c) adverbial modifiers

Figure6: “Unlik e CoordinatedPhrases”

caseswhenthewh constituentis in a non-whposi-tion, we needtheoppositeeffect: a WHNP (or wh-nounPOStag)is allowedto projectup to anNP.

3.3 Unlik eCoordinated Phrases(UCP)This is the expressionusedin the PTB to denotecoordinatedphrasesin which the coordinatedcon-stituentsarenotof thesamesyntacticcategory. Therationalefor the existenceof suchconstructionsisthat the coordinatedconstituentsarealternative re-alizationsof the samegrammaticalfunction withrespectto a lexical head. In Figure 6.a, both anounandanadjectiveareallowedto modify anothernoun,andthereforethey canbeconjoinedwhile re-alizing thatfunction. Two othercommoncasesare:coordinationof predicatesin copularconstructions(Figure6.b)andadverbialmodification(Figure6.c).

Wedealwith theproblemasfollows. First,weal-low for aUCPto beextractedasanargumentwhentheheadis a verb andtheUCP is marked predica-tive (PRDfunction tag) in the trainingexample;orwhenever the headis seento have an obligatoryargument requirement(e.g., prepositions: “Theycome from ((NP the house)and (PP behind thetree))”). Second,a UCP is allowed to modify (ad-join to) mostof thenodes,accordingto evidenceinthecorpusandcommonsense(in thefirst andthirdexamplesabove we hadNP andVP modification).With respectto thehosttree,whenattachedasanar-gumentthey aretreatedlikeany othernon-terminal:a substitutionnode.Theleft treein Figure7 shows

,.-/ / /10002434-657,8-:9 2434-/ / / / 0000,.,;5<3�3>=@?A?B5Figure7: Extractedtreesfor UCP

thecasewheretheUCPis treatedasa modifier. Infactthetreesarebothfor theexamplein Figure6.a.Notice that the tree is non-lexicalized to avoid ef-fectsof sparseness.TheUCPis thenexpandedasintheright treein Figure7: aninitial treeanchoredbytheconjunction(thetreeattacheseitherto atreeliketheonein the left or asa trueargument– the latterwouldbethecasefor theexamplein Figure6.b).

Now, the caveats. First, we aregiving the UCPthe statusof an independentnon-terminal,as if ithadsomeintrinsic categorial significance(asasyn-tacticprojection).Theassumptionof independenceof expansion,that for context-free grammarsis in-herentto eachnon-terminal,in TAGs is further re-stricted to the substitutionnodes. For example,whenanNP appearsassubstitutionnode,in a sub-jector objectposition,or asanargumentof aprepo-sition or a genitive marker, we arestatingthat anypossibleexpansionfor theNPis licensedthere.Thesamehappensfor otherlabelsin argumentpositionsaswell. While thatis anovergeneratingassumption(e.g. theexpletive “there” cannotbetherealizationof anNPin objectposition),it is generallytrue.Forthe UCP, however, we know that its expansionisin fact strongly dependenton where the substitu-tion nodeis, as we have arguedbefore. In fact itis lexically dependent(cf. “I know ((the problem)and(thatthereis nosolutionto it))”, wherethecon-junctsarelicensedby thesubcategorizationsof theverb “know”). On theotherhand,it doesnot seemreasonableto expandthe UCP nodeat the hostingtree– a crossproductexplosion.A possibleway ofalleviating this effect could be to expandonly theauxiliary trees(a UCP modifying a VP is distinctfrom a UCP modifying an NP, andmoreover theyareindependentof lexical items).But for trueargu-mentpositionsthereseemsto beno clearsolution.

Second,theoddityof theUCPasalabelbecomesapparentonceagainwhen thereare multiple con-juncts,asin Figure8: it is enoughfor oneof themtobedistinct to turn theentireconstituentinto a UCP.Recursive decompositionin the grammarin thesesituationsclearlyleadsto somenon-standardtrees.

Finally, andmorecrucially, we have omittedonecasein our discussion:the casein which the UCP

39

Page 44: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(NP (UCP (JJ electronic)(, ,)(NN computer)(CC and)(NN building))

(NNS products))

Figure8: UCPwith multiple conjuncts

(S (NP-SBJ-1 The Series 1989 B bonds)(VP (VBP are)

(VP (VBN rated)(S *-1 double-A))))

(S (NP-SBJ-1 The Series 1989 B bonds)(VP (VBP are)

(UCP-PRD (ADJP-PRD (JJ uninsured))(CC and)(VP (VBN rated)

(S *-1 double-A)))))

Figure9: UCPinvolving VP argumentof thecopula

is thenaturalhead-childof somenode.Undersomeaccountsof grammardevelopmentthis never hap-pens:we have observed that UCP doesnot appearasheadchild in the accountwherethe headis thesyntacticheadof a node. We have not alwaysfol-lowedthis rule. With respectto theVP head,sofarwe have followed onemajor tendency in the com-putationalimplementationof lexicalizedgrammars,accordingtowhich lexical verbsarepreferedtoaux-iliary verbsto headtheVP. Now, considerthepairof sentencesin Figure9.

Underthe lexical verbparadigm,in thefirst sen-tencethe derivation would startwith an initial treeanchoredby thepastparticipleverb (“rated”). Butthenwe have an interestingproblemin the secondsentence,for which we do not currentlyhave a neatsolution. Following Xia’s samplesettingsof Lex-Tract parameters,in thesecasesthe extraction isrescuedby switchingto theotherparadigm:theini-tial treeis extractedanchoredby theauxiliary verbwith a UCP argument,andtheVP is acceptedasapossibleconjunct. A systematicmove to the syn-tactic head paradigm,which we may indeedtry,would have importantconsequencesin the localityassumptionsfor thegrammardevelopment.

3.4 VP topicalizationAnother problem with the lexical verb paradigm(seealso discussionunderUCP above) is the VPtopicalizationasin the sentencein Figure10. Thesolution currently adopted(again, inherited from

(SINV (ADVP (RB Also))(VP-TPC-2 (VBN excluded)

(NP (-NONE- *-1)))(VP (MD will)

(VP (VB be)(VP (-NONE- *T*-2))))

(NP-SBJ-1 investments in ...))

Figure10: VP topicalization

(S (NP-SBJ (NNP Congress))(VP (MD could)

(VP (VB pass)(ADVP-MNR (RB quickly))(NP (NP (DT a)

(‘‘ ‘‘)(JJ clean)(’’ ’’)(NN bill))

(VP (VBG containing)(ADVP (JJ only))(NP ... ))))))

C/ / / 000,8-�5 D -/ / 00D =7,.-E5Figure11: Theextrapositionproblem

Xia’s samplesettings)is asabove: theparadigmisswitchedandtheauxiliary verb(“be”) is chosenastheanchorof theinitial tree.

3.5 Extraposition and Verb Subcategorization

One of the key designprinciples that have beenguidinggrammardevelopmentwith TAGsis to keepverb argumentsas substitutionslots local to thetreeanchoredby the verb. It is widely known thatthe PennTreebankdoesnot distinguishverb ob-jectsfrom adjuncts.Sosomesortsof heuristicsareneededto decide,amongthecandidates,which areto betakenasarguments(KinyonandProlo,2002);the rest is extractedasseparateVP modifier trees.However, this step is not enoughfor the treestocorrectly reflect verb subcategorizations. The oc-currenceof discontinuousarguments,frequentlyex-plainedasargumentextraposition(theargumentisraisedpastthe adjunct)createsa problem. In thesentencein Figure11theverb“pass”shouldanchora treewith oneNP object.

However in sucha treeit would beimpossibletoadjointhetreefor theinterveningADVP “quickly”asa VP modifier andstill have it betweentheverbandtheNP.9 LexTractthenwould insteadextractan

9A strikinguseof sisteradjunctionin (Chiang,2000)is ex-actly theelegantway it solvesthis problem:thenon-argumenttreecanbeadjoinedontoanode(say, VP), positioningitself inbetweentheVP’schildren,which is notpossiblewith TAGs.

40

Page 45: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(NP (NP the 3 billion New Zealand dollars)(PRN (-LRB- -LRB-)

(NP US$ 1.76 billion *U*)(-RRB- -RRB-)))

a)A parentheticalNP attachedto anotherNP

(S (NP-SBJ The total relationship)(PRN (, ,)

(SBAR-ADV as Mr. Lee sees it)(, ,))

(VP (VBZ is) ...))

b) A parentheticalS betweensubjectandverb

Figure12: Parentheticals

intransitive treefor the VB “pass”, onto which theADVP modifiertreewouldadjoin.Thesecondodd-ity is thattheNPobjectwouldalsobeextractedasaVP modifiertree. In a nutshell,objectsin extractedtreesarerestrictedto thosewhicharenotextraposedandhencethetreesmaynot truly reflecttheproperdomainof locality. Oneview is that thesetof treesfor acertainsubcategorizationframewould includethesedegeneratecases.LexTract hasan option toallow limited discontinuity, i.e.,a non-argumentse-quencebetweentheverbandthefirst object(but notbetweentwo objects). The non-argumentswouldthenbe adjoinedto the V node.10 So far we haveusedonly thelatteralternative.

It is worth mentioningtwo othercasesof extra-position.Subjectextrapositionis handledby havingtheextraposedsubject,usuallyasententialform,ad-join at theVP of which it is thelogical subject(theoriginal positionis still occupiedby anNP with theexpletive pronoun“it”). Relative clauseextraposi-tion is modeledby a relative clausetree,only it ad-joins ataVP, insteadof at anNP asis usual.

3.6 ParentheticalsParentheticalexpressionsare ubiquitous in lan-guage:they mayappearalmosteverywherein asen-tenceandcanbeof almostany category (Fig. 12).

We modelthemasadjoining,eitherto the left orright of the constituentthey aredominatedby, de-pendingonwhetherthey areto theleft or rightof theheadchild of theparent’s node. Occasionallysuchtreescanalsobeinitial. Therespective treesfor theexamplesof Figure12 aredrawn in Figure13. It

10Of course,althoughthesolutioncoversmostof theoccur-rences,andapartof any linguistic concern,therearestill un-coveredcases,e.g.,whena parentheticalexpressionintervenesbetweenthefirst andthesecondargument.

,8-/ / 00,.-$9F-4G4,,.-�5 C/ / 00-�G4,CAH:I G65 C 9Figure13: Extractedtreesfor parentheticals

is always the casethat the label PRN dominatesasinglesubstitutionnode.Whenever thiswasnot thecasein the trainingcorpus,heuristicsbasedon ob-servationwereusedto enforcethat,by insertinganappropriatemissingnode.

3.7 Projection labels

LexTractextractstreeswith no concernfor theap-propriateprojective structureof constituentswhennot explicitly marked in thePTB. Figure14 showstwo examplesof NP modificationwherethemodi-fiersaresinglelexical items.Theextractedmodifiertrees,shown on the right, do not have the projec-tion for the modifiersJJR“stronger” andthe NNP“October” (which shouldbe,respectively, anADJPandanNP).Thatis so,becausethosenodesarenotfoundin theannotation.

(NP (DT a)(JJR stronger)(NN argument))

(NP-SBJ-1 (NNP October)(NN weather))

,8-/ / 00?J?G>=7,.-$9,.-/ / /K000,8,.-�=7,8-:9

Figure14: Simplemodificationannotationandex-tractedtrees

However, if themodifiersarecomplex, that is, ifthemodifiersarethemselvesmodified,thePTB in-sertstheir respective projections,andthereforetheyappearin theextractedtrees,asshown in Figure15.

Thereseemsto be no reasonfor the two pairsof extractedtreesto be different. Much of this iscausedby the acknowledgedflatnessin the PennTreebankannotation.Thatsaid,thetreeslike thosein thesecondpair shouldbepreferred.Theprojec-tion node(ADJP or NP) is understoodto be domi-natingits headeven whenthereis no furthermod-ification, andit shouldbe a concernof a goodex-tractionprocessto insertthe missingnodeinto thegrammar. SinceLexTract do not allow us to spec-

41

Page 46: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

(NP (DT an)(ADJP (RB even)

(JJR stronger))(NN argument))

(NP-SBJ-1 (NP (JJ late)(NNP October))

(NN weather))

,8-/ / / 000I>L ?-?A?MG>= ,8-:9,.-/ / / 000,.-,8,.-�= ,8-:9

Figure 15: Complex modification annotationandextractedtrees

ify for the insertionof “obligatory” projectionswehadto accomplishthis throughasomewhatcompli-catedpost-processingstepusinga projectiontable.Someof our current projectionsare: nouns,per-sonalpronounsandtheexistentialexpletive to NP;adjectivesto ADJP;adverbsto ADVP; sentencesei-therto SBAR (S,SINV) or to SBARQ (SQ);Cardi-nals(CD) to QuantifierPhrases(QP) which them-selvesproject to NP. Notice that not all categoriesare forcefully projected. For instance,verbs arenot, allowing for simple auxiliary extraction. INis also not projecteddue to its double role as PPhead (true preposition)and subordinateconjunc-tion, whichshouldprojectontoSBARs.

4 ConclusionWe discussedanexperimentin grammarextractionfrom corporawith focuson problemsarisingwhiletrying to give anadequateaccountfor naturallyoc-curingphenomena.Withoutbeingexhaustive in ourlist, we expectto have brougtsomeattentionto theneedto discusssolutionsfor themwhichareasrea-sonableaspossiblegiventhecurrentstate-of-the-artof the linguistic research,computationalgrammardevelopmentand automaticextraction, and giventhecurrentcorpusresourcesatourdisposition.

ACKNOWLEDGEMENTS: Thanks to ToniaBleam, Erwin Chan, AlexandraKinyon, RashmiPrasad,BeatriceSantorini,Fei Xia and the XTAGGroupfor valuablediscussionsalongtherealizationof this work and/orcommentson this paperor re-latedmaterial.

ReferencesJoanBresnanandJaneGrimshaw. 1978. Thesyn-

taxof freerelativesin english.LinguisticInquiry,9(3):331–391.

JohnChenandK. Vijay-Shanker. 2000.Automatedextractionof TAGs from the PennTreebank.InProceedingsof the6thInternationalWorkshoponParsingTechnologies, Trento,Italy.

David Chiang. 2000. Statisticalparsingwith anautomatically-extracted Tree Adjoining Gram-mar. In Proceedingsof the38thAnnualMeetingof theAssociationfor ComputationalLinguistics,HongKong,China.

RobertFrank.2002.PhraseStructure Compositionand SyntacticDependencies. MIT Press,Cam-bridge,MA, USA.

AnnekeGroosandHenkvonRiemsdijk.1979.Thematchingeffectsin freerelatives: a parameterofcoregrammar. In Theoryof Markednessin Gen-erative Grammar. ScuolaNormaleSuperiorediPisa,Italy.

Aravind K. Joshiand YvesSchabes.1997. Tree-Adjoining Grammars. In Handbookof FormalLanguages, volume 3, pages69–123.Springer-Verlag,Berlin.

Alexandra Kinyon and Carlos A. Prolo. 2002.Identifying verb argumentsand their syntacticfunction in the PennTreebank. In Proc. of theThird LREC, pages1982–87,LasPalmas,Spain.

Mitchell Marcus,BeatriceSantorini,andMary AnnMarcinkiewicz. 1993.Building alargeannotatedcorpusof English: ThePennTreebank.Compu-tational Linguistics, 19(2):313–330.

Mitchell Marcus, Grace Kim, Mary AnnMarcinkiewicz, Robert MacIntyre, Ann Bies,Mark Ferguson,Karen Katz, and Britta Schas-berger. 1994. The PennTreebank:Annotatingpredicateargument structure. In Proceedingsof the 1994 Human Language TechnologyWorkshop.

CarlosA. Prolo. 2002.LR parsingfor TreeAdjoin-ing Grammarsandits applicationto corpus-basednaturallanguageparsing.Ph.D.DissertationPro-posal,Universityof Pennsylvania.

Fei Xia. 1999. Extracting tree adjoining gram-marsfrom bracketedcorpora.In Proceedingsofthe5thNatural LanguageProcessingPacificRimSymposium(NLPRS-99), Beijing, China.

Fei Xia. 2001. Investigatingthe Relationshipbe-tweenGrammarsandTreebanksfor Natural Lan-guages. Ph.D. thesis,Departmentof ComputerandInformationScience,Un. of Pennsylvania.

42

Page 47: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

A Classificationof Grammar DevelopmentStrategies

Alexandra Kiny on Carlos A. ProloComputerandInformationScienceDepartment

Universityof PennsylvaniaSuite400A,3401WalnutStreet

Philadelphia,PA, USA, 19104-6228Nkinyon,proloO @linc.cis.upenn.edu

Abstract

In this paper, we proposea classificationof gram-mar developmentstrategiesaccordingto two crite-ria : hand-written versusautomatically acquiredgrammars,and grammarsbasedon a low versushigh level of syntactic abstraction. Ourclassifica-tion yields four typesof grammars.For eachtype,we discussimplementationandevaluationissues.

1 Intr oduction: Four grammardevelopmentstrategies

Thereareseveralpotentialstrategiesto build wide-coveragegrammars,thereforethere is a needforclassifyingthesevariousstrategies. In this paper,we proposea classificationof grammardevelop-mentstrategiesaccordingto two criteria:P Hand-crafted versus Automatically ac-

quir ed grammarsP Grammarsbasedon a low versushigh level ofsyntacticabstraction.

As summarizedin table 1, our classificationyields four typesof grammars,which we call re-spectively typeA, B, C andD.

Of thesefour types,threehave alreadybeenim-plementedto develop wide-coveragegrammarsforEnglishwithin theXtag project,andan implemen-tation of the fourth type is underway 1. Most ofourexamplesarebasedonthedevelopmentof widecoverageTreeAdjoining Grammars(TAG), but it isimportantto note that the classificationis relevantwithin otherlinguistic frameworks aswell (HPSG,GPSG,LFG etc.)andis helpful to discussportabil-ity amongseveralsyntacticframeworks.

We devotea sectionfor eachtypeof grammarinour classification. We discussthe advantagesanddrawbacksof eachapproach,andespeciallyfocus

1We do not discusshereshallow-parsingapproaches,butonly f ull grammardevelopment.Dueto spacelimitations,wedo not introducetheTAG formalismandreferto (Joshi,1987)for anintroduction.

on how eachtype performsw.r.t. grammarcov-erage,linguistic adequacy, maintenance,over- andunder- generationaswell asto portability to othersyntacticframeworks. We discussgrammarrepli-cationasa meanto comparetheseapproaches.Fi-nally, we argue that the fourth type, which is cur-rently beingimplemented,exhibits betterdevelop-mentproperties.

2 TYPE A Grammars: hand-craftedThelimitationsof TypeA grammars(hand-crafted)are well known : although linguistically moti-vated,developing and maintaininga totally hand-craftedgrammaris a challenging(perhapsunreal-istic ?) task. Sucha large hand-craftedgrammarfor TAGs is describedfor English in (XTAG Re-searchGroup, 2001). Smallerhand-craftedgram-marsfor TAGs have beendevelopedfor other lan-guages(e.g. French(Abeille, 1991)),with similarproblems.Of course,thelimitationsof hand-craftedgrammarsare not specific to the TAG framework(seee.g.(ClementandKinyon,2001)for LFG).

2.1 CoverageissuesThe Xtag grammarfor English, which is freelydownloadablefrom the projecthomepage2 (alongwith tools suchas a parserand an extensive doc-umentation),hasbeenunderconstantdevelopmentfor approximately15years.It consistsof morethan1200elementarytrees(1000for verbs)andhasbeentestedonrealtext andtestsuites.For instance,(Do-ranet al., 1994)reportthat61%of 1367grammat-ical sentencesfrom the TSNLP test-suite(Lehmanandal, 1996)wereparsedwith an early versionofthe grammar. More recently, (Prasadand Sarkar,2000) evaluatedthe coverageof the grammaron”the weathercorpus”,which containedrathercom-plex sentenceswith an averagelengthof 20 wordsper sentence,as well as on the ”CSLI LKB testsuite” (Copestake, 1999). In addition, in order to

2http://www.cis.upenn.edu/xtag/

43

Page 48: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

High levelof syntacticab-straction

Low level of syntacticab-straction

Hand-crafted TypeA: TypeC:Traditional hand-craftedgrammars

Hand-craftedlevel of syn-tacticabstractionAutomatically generatedgrammars

Automaticallyacquired TypeB: TypeD:Traditional treebank ex-tractedgrammars

Automatically acquiredlevel of syntacticabstrac-tionAutomatically generatedgrammar

Table1: A classificationof grammars

evaluatethe rangeof syntacticphenomenacoveredby the Xtag grammar, an internal test-suitewhichcontainsall the example sentences(grammaticaland ungrammatical)from the continually updateddocumentationof the grammaris distributed withthegrammar. (PrasadandSarkar, 2000)arguethatconstantevaluationis usefulnot only to getanideaof thecoverageof a grammar, but alsoasa way tocontinuouslyimprove andenrichthegrammar3.

Parsing failures were due, amongother things,to POSerrors,missinglexical items,missingtrees(i.e. grammarrules), featureclashes,bad lexicongrammarinteraction(e.g.lexical itemanchoringthewrongtree(s))etc.

2.2 Maintenanceissues

As a hand-craftedgrammargrows , consistency is-suesariseand one then needsto develop mainte-nancetools. (Sarkarand Wintner, 1999) describesucha maintenancetool for the Xtag grammarforEnglish, which aims at identifying problemssuchas typographicalerrors (e.g. a typo in a featurecanprevent unificationat parsetime andhurt per-formance),undocumentedfeatures(featuresfromolder versionsof the grammar, that no longer ex-ist), type-errors(e.g.Englishverbnodesshouldnotbe assigneda genderfeature),etc. But even withsuchmaintenancetools, coverage,consistency andmaintenanceissuesstill remain.

3For instance,at first, Xtag parsedonly 20% of the sen-tencesin theweathercorpusbecausethiscorpuscontainedfre-quentfree relative constructionsnot handledby the grammar.After augmentingthegrammar, 89.6%of thesentencesdid geta parse.

2.3 Ar ehand-crafted grammars useful ?

Somedegree of automationin grammardevelop-ment is unavoidablefor any real world application: small andeven medium-sizehand-craftedgram-mararenotusefulfor practicalapplicationsbecauseof their limited coverage,but largergrammarsgiveway to maintenanceissues. However, despitetheproblemsof coverageandmaintenanceencounteredwith hand-craftedgrammars,suchexperimentsareinvaluablefrom a linguistic point of view. In par-ticular, the Xtag grammarfor English comeswitha very detaileddocumentation,which hasprovedextremelyhelpful to devise increasinglyautomatedapproachesto grammardevelopment(seesectionsbelow) 4.

3 TYPE B Grammars: Automaticallyextracted

To remedysomeof theseproblems,Type B gram-mars(i.e. automaticallyacquired,mostly from an-notatedcorpora)have beendeveloped.For instance(Chiang,2000),(Xia, 2001)(Chen,2001)all auto-matically acquirelarge TAGs for Englishfrom thePennTreebank(Marcuset al., 1993).However, de-spite an improvementin coverage,new problemsarisewith this typeof grammars: availability of an-notateddatawhich is large enoughto avoid sparsedataproblems,possiblelackof linguistic adequacy,extractionof potentially unreasonablylarge gram-mars(slowsdown parsingandincreasesambiguity),

4Perhapsfully hand-craftedgrammarscanbeusedin prac-tice on limited domains,e.g. the weathercorpus.However, adegreeof automationis usefuleven in thosecases,if only toinsureconsistency andavoid somemaintenanceproblems.

44

Page 49: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

lackof domainandframework independence(e.g.agrammarextractedfrom thePennTreebankwill re-flect thelinguistic choicesandtheannotationerrorsmadewhenannotatingthetreebank).

We give two examplesof problemsencounteredwhenautomaticallyextractingTAG grammars:Theextractionof a wrongdomainof locality; And Theproblemof sparse-dataregardingthe integrationofthelexicon with thegrammar.

3.1 Wr ongdomain of locality

Long distancedependenciesare difficult to detectaccuratelyin annotatedcorpora,even when suchdependenciescan be adequatelymodeledby thegrammarframework usedfor extraction (which isthecasefor TAGs,but not for instancefor ContextFreeGrammars).For example,(Xia, 2001)extractstwo elementarytreesfrom asentencesuchasWhichdog doesHillary Clinton thinkthatChelseaprefers.Thesetreesareshown on figure 1. Unfortunately,becauseof the potentiallyunboundeddependency,the two treesexhibit an incorrectdomainof local-ity: theWh-extractedelementendsup in thewrongelementarytree,asanargumentof ”think”, insteadof asanargumentof ”prefer” 5QR R R R R SSSSSTAUWVYX Z\[ ]

(Which dog)

QR R R R SSSSTJU>](Hillary) ^ UR R SS^`_(think)

Q ]QR R R R SSSSTJU>]

(Chelsea) ^ UR R R SSS^`_(prefers)

TJUbabc dfehgiFigure1: Extractionof thewrongdomainof locality

This problemis not specificto TAGs,andwouldtranslatein otherframeworks into theextractionofthe”wrong” dependency structure6.

3.2 Sparsedata for lexicon-grammarintegration

Existing extractionalgorithmsfor TAGs acquireafully lexicalizedgrammar. A TAG grammarmaybeviewedasconsistingof two components:ontheone

5Someextractionalgorithmssuchasthoseof (Chen,2001)or (Chiang,2000)doretrievetheright therightdomainof local-ity for thisspecificexample,but doextractadomainof localitywhich is incorrectin someothercases.

6Onecanarguethat theproblemdoesnot appearwhenus-ing simpleCFGs,and/orthatthisproblemis only of interesttolinguists. A counter-argumentis that linguistic adequacy of agrammar, whetherextractedor not,DOESmatter. An extremecaricatureto illustratethis point : the context freegrammarSj S word k word allows oneto robustly andunambiguouslyparseany text, but is not very usefulfor any furtherNLP.

hand“tree templates”andon the otherhanda lex-icon which indicateswhich treetemplate(s)shouldbeassociatedto eachlexical item 7.

Supposethe following three sentencesare en-counteredin thetrainingdata:

1. Peterwatchesthestars2. Mary eatstheapple3. WhatdoesPeterwatch ?From thesethreesentences,two tree templates

will be correctly acquired,as shown on figure 2 :The first onecovers the canonicalorderof the re-alization of argumentsfor sentences1 and 2, thesecondcoversthecaseof aWh-extractedobjectforsentence3. Concerningtheinteractionbetweenthelexiconandthegrammarrules,thefactthat“watch”shouldselectbothtreeswill beaccuratelydetected.However, thefact that“eat” shouldalsoselectbothtreeswill be missedsince“eat” hasnot beenen-counteredin aWh-extractedObjectconstruction.

Anchorseatandwatchlm m m nnno�pbqYr s�pm m nnsutvo�p�w`rAnchorsonly watchlm m m m m nnnnno�pyxfz|{f}�~�r lm m m nnno4pu�Jr supm m nnsJt o�p w�

Figure 2: Correct templates, but incompletelexicon-grammarinterface

A level of syntacticabstractionis missing: in thiscase,thenotionof subcategory frame.This is espe-cially noticeablewithin the TAG framework fromthe fact that in a TAG hand-craftedgrammarthegrammarrulesaregroupedinto “treefamilies”,withonefamily for eachsubcategorizationframe(tran-sitive, intransitive, ditransitive, etc.), whereasau-tomaticallyextractedTAGs do not currentlygrouptreesinto families.4 TYPE C GrammarsTo remedythe lack of coverageand maintenanceproblemslinked to hand-craftedgrammars,aswellasthelackof generalizationandlinguisticadequacyof automaticallyextractedgrammars,new syntac-tic levels of abstractionare defined. In the con-text of TAGs,onecancite thenotionof MetaRules(Becker, 2000), (Prolo, 2002)8, and the notion ofMetaGrammar(Candito,1996),(Xia, 2001).

7This subdivision avoids an combinatoricexplosionin thenumberof rulesif thegrammarwasfully lexicalized

8For otherMetaRulebasedapproachesbasedon theDATRformalism,see(Carroll etal., 2000)or (Evansetal., 2000)

45

Page 50: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

4.1 MetaRules

A MetaRuleworks as a pattern-matchingtool ontrees. It takes as input an elementarytree andoutputsa new, generallymore complex, elemen-tary tree. Therefore, in order to createa TAG,one can start from one canonicalelementarytreefor eachsubcategorizationframeanda finite num-ber of MetaRuleswhich model syntactictransfor-mations(e.g. passive, wh-questionsetc) and au-tomatically generatea full-size grammar. (Prolo,2002)startedfrom 57elementarytreesand21hand-craftedMetaRules,andre-generatedthe verb treesof the hand-craftedXtag grammarfor Englishde-scribedin theprevioussection.

The replicationof the hand-craftedgrammarforEnglish,usingaMetaRuletool, presentsinterestingaspects: it allows to directly comparethe two ap-proaches.Sometreesgeneratedby (Prolo, 2002)werenot in thehand-craftedgrammar(e.g. variousorderingsof “by phrasepassives”) while someoth-ersthatwerein thehand-craftedgrammarwerenotgeneratedby theMetaRules9. This replicationpro-cessmakesit possible,with detailedscrutiny of theresults,to :P Identify what shouldbe considerasunder- or

over- generationof theMetaRuletool.P Identify what should be consideredto beunder- or over- generationof thehand-craftedgrammar.

Thus,grammarreplicationtasksmake it possibleto improve boththehand-craftedandtheMetaRulegeneratedgrammars.4.2 MetaGrammarsAnotherpossibleapproachfor compactandabstractgrammarencodingis theMetaGrammar(MG), ini-tially developedby (Candito,1996). The ideais tocompactlinguistic information thanksto an addi-tional layerof linguisticdescription,which imposesageneralorganizationfor syntacticinformationin athree-dimensionalhierarchy:P Dimension1: initial subcategorizationP Dimension2: valency alternationsandredistri-

bution of functionsP Dimension3: surfacerealizationof arguments.

Eachterminal classin dimension1 describesapossibleinitial subcategorization(i.e. a TAG treefamily). Each terminal classin dimension2 de-scribesa list of orderedredistributionsof functions(e.g. it allows to add an argumentfor causatives,

9Due to spacelimitations, we refer to (Prolo, 2002) for adetaileddiscussion.

to eraseone for passive with no agents...). Eachterminal class in dimension3 representsthe sur-facerealizationof a surfacefunction (ex: declaresif a direct-objectis pronominalized,wh-extracted,etc.). Each classin the hierarchycorrespondstothe partial descriptionof a tree(RogersandVijay-Shanker, 1994). A TAG elementarytree is gener-atedby inheriting from exactly one terminal classfrom dimension1, oneterminalclassfrom dimen-sion 2, and n terminal classesfrom dimension3(where n is the numberof argumentsof the ele-mentarytreebeinggenerated).For instancetheele-mentarytreefor ”Parqui seraaccompagneeMarie”(By whomwill Mary beaccompanied)is generatedby inheriting from transitivein dimension1, fromimpersonal-passivein dimension 2 and subject-nominal-inverted for its subject and questioned-object for its object in dimension3. This compactrepresentationallows one to generatea 5000 treegrammarfrom a hand-craftedhierarchyof a fewdozensof nodes,esp.sincenodesareexplicitly de-finedonly for simple syntacticphenomena10. TheMG wasusedto developa wide-coveragegrammarfor French(Abeille etal., 1999).It wasalsousedtodevelopamedium-sizegrammarfor Italian,aswellasagenerationgrammarfor German(Gerdes,2002)usingthenewly availableimplementationdescribedin (Gaiffe et al., 2002). A similar MetaGrammarapproachhasbeendescribedin (Xia, 2001)for En-glish 11.

4.3 MetaGrammars versusMetaRules: whichis best?

It would be desirableto have a way of compar-ing the resultsof theMetaGrammarapproachwiththatof theMetaRuleapproach.Unfortunately, thisis not possiblebecauseso far noneof the two ap-proacheshavebeenusedwithin thesameproject(s).Therefore,in orderto have a bettercomparisonbe-tweenthesetwo approaches,we have starteda sec-ond replicationof the Xtag grammarfor English,this time using a MG. This replicationshouldal-low us to make a direct comparisonbetweenthehand-craftedgrammar, thegrammargeneratedwithMetaRulesandthegrammargeneratedwith aMG.

For this replicationtask,we usethemorerecentimplementationpresentedin (Gaiffe et al., 2002)becausetheir tool :

10Nodesfor complex syntacticphenomenaaregeneratedbyautomaticcrossingsof nodesfor simplephenomena

11but that particularwork did not attemptto replicatetheXtag grammar, andthusthegeneratedgrammaris not directlycomparableto thehand-craftedversionof thegrammar.

46

Page 51: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

P Is freely available 12, portable (java), wellmaintainedandincludesa GraphicalUserIn-terface.P Outputsa standardizedXML format13P Is flexible (onecanhave more than3 dimen-sionsin the hierarchy)andstrictly monotonicw.r.t. thetreesbuiltP Supports“Hypertags”, i.e. eachelementarytreein thegrammaris associatedwith afeaturestructurewhich describesits salientlinguisticproperties14.

In the (Gaiffe et al., 2002)implementation,eachclassin theMG hierarchycanspecify:P Its SuperClasse(s)P A Featurestructure(i.e. Hypertag)which cap-

tures the salient linguistic characteristicsofthatclass.P WhattheclassneedsandprovidesP A setaquasi-nodesP Constraintsbetweenquasi-nodes(father, dom-inates,precedes,equals)P traditionalfeatureequationsfor agreement.

The MG tool automaticallycrossesthe nodesinthehierarchy, looking to create“balanced”classes,thatis classesthatdonotneednorprovideanything.From thesebalancedterminal classes,elementarytreesaregenerated.Figure3 shows how a canon-ical transitive treeis automaticallygeneratedfrom 3hand-writtenclassesandthe quasi-treesassociatedto theseclasses15.4.4 Advantagesand drawbacksof TYPE C

grammarsIt is often assumedthat Metaruleand MetaGram-mar approachesexhibit some of the advantagesof hand-craftedgrammars(linguistic relevance)aswell assomeof theadvantagesof automaticallyex-tractedgrammars(wide-coverage),aswell aseasiermaintenanceandbettercoherence.However, as ispointedout in (Barrier et al., 2000),grammarde-velopmentbasedon hand-craftedlevelsof abstrac-tion give rise to new problemswhile not necessar-ily solvingall theold problems:Althoughtheauto-maticgenerationof thegrammarinsuressomelevel

12http://www.loria.fr/equipes/led/outils/mgc/mgc.html13Seehttp://atoll.inria.fr/clerger/tag20.dtd,xmlfor morede-

tailson formatstandardizationefforts for TAG relatedtools.14The ideaof “featurization” is very usefulfor applications

suchas text generation,supertagging(Kinyon, 2002), and isespeciallyrelevant for theautomaticacquisitionof a MG (seesection5)

15Thisexampleis of courseasimplification:for sakeof clar-ity it doesnot reflectthecomplex structureof our real”hierar-chy”.

Figure3: Generatingacanonicaltransitive treewitha MetaGrammarof 3 classes: � standsfor ”f atherof”, � for ”precedes”,� for anchornodesand � forsubstitutionnodes.

of consistency, problemsariseif mistakesaremadewhile hand-craftingtheabstractlevel (hierarchyorMetaRules)from which the grammaris automati-cally generated.Thisproblemis actuallymoreseri-ousthanwith simplehand-craftedgrammars,sincean error in onenodewill affect ALL treesthat in-herit from this node. Furthermore,a large por-tion of the generatedgrammarcovers raresyntac-tic phenomenathatarenot encounteredin practice,whichunnecessarilyaugmentsthesizeof theresult-ing grammars,increasesambiguitywhile notsignif-icantly improving coverage16. One crucial prob-lem is that despitethe automaticgenerationof thegrammar(which easesmaintenance),the interfacebetweenlexicon andgrammaris still mainly man-

16For instance,the5000treegrammarfor Frenchparses80%of (simple)TSNLP sentences,anddoesnot parsenewspapertext, whereasthe1200treehand-craftedXtaggrammarfor En-glish does.Basically, insteadof solvingbothunder-generationandover-generationproblems,a hand-craftedabstractlevel ofsyntacticencodingrunstherisk of increasingboth

47

Page 52: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ually maintained(and of courseone of the majorsourcesof parsingfailuresis dueto missingor erro-neouslexical entries).

5 TYPE D GrammarsHowever, the main potentialadvantageof suchanabstractlevel of syntacticrepresentationis frame-work independence.We arguethat themaindraw-backsof an abstractlevel of syntacticrepresenta-tion (over-generation,propagationof manualerrorsto generatedtrees,interfacewith the lexicon) maybesolved if this abstractlevel is acquiredautomat-ically insteadof being hand-crafted. Other prob-lemssuchassparsedataproblemsarealsohandledby sucha level of abstraction17. This correspondsto type D in our classification. A preliminaryde-scription of this work, which consistin automati-cally extractingthehierarchynodesof aMetaGram-mar from the PennTreebank(i.e. a high level ofsyntacticabstraction)maybefoundin (Kinyon andProlo, 2002). The underlyingidea is that a lot ofabstractframework independentsyntacticinforma-tion is implicitly presentin thetreebank,andhastobe retrieved. This includes: subcategorizationin-formation,potentialvalency alternations(e.g. pas-sivesaredetectedby amorphologicalmarkeronthePOSof the verb, by the presenceof an NP-Object”trace”, and possiblyby the presenceof a Prepo-sitional phraseintroducedby ”by”, and markedas ”logical-subject”),andrealizationof arguments(e.g. Wh-extractionsare noticedby the presenceof a Wh constituent,co-indexed with a trace). Inorder to retrieve this information, we have exam-ined all the possibletag combinationsof the PennTreebank2 annotationstyle, andhave determinedfor eachcombination,dependingon its locationintheannotatedtreewhetherit wasanargument(op-tional or compulsory)or a modifier. We mappedeachargumentto a syntacticfunction 18. This al-lowed us to extract fine-grainedsubcategorizationframesfor eachverb in the treebank.Eachsubcat-egorizationframeis storedasa finite numberof fi-nal classesusingthe(Gaiffe et al., 2002)MG tool :oneclassfor eachsubcategorizationframe(dimen-sion1 in Candito’s terminology),andoneclassfor

17As discussedin section3, if oneseeseat in thedata,andoneseessomeothertransitiveverbwith aWh extractedobject,the elementarytreefor “What doesJ. eat” is correctlygener-ated,even if eat hasnever beenencounteredin sucha con-structionin thedata,which is not the casewith theautomaticextractionof traditional lexicalizedgrammars

18We usethe following functions: subject,predicative, di-rectobject,secondobject,indirectobject,LocDir object.

eachfunctionrealization(dimension3 in Candito’sterminology). The sametechniqueis usedto ac-quirethevalency alternationfor eachverb,andnon-canonicalsyntacticrealizationsof verb arguments(Wh extractionsetc...). This amountsto extract-ing ”hypertags”(Kinyon, 2000)from the treebank,transformingtheseHypertagsinto aMetaGrammar,andautomaticallygeneratinga TAG from theMG.An exampleof extractionmaybeseenon figure4 :exposeappearsherein a reduced-relative construc-tion. However, from thetraceoccupying thecanon-ical positionof a directobject,theprogramretrivesthecorrectsubcategorizationframe(i.e. treefamily)for this verb. Hence,just this occurenceof exposecorrectly extractsthe MG nodesfrom which boththe”canonicaltree”andthe”Reducedrelative tree”will be generated.If onewas extracting a simpletypeB grammar, thecanonicaltreewouldnotbere-trievedin thisexample.

Input Sentence :-------------(NP (NP (DT a)

(NN group) )(PP (IN of)

(NP (NP (NNS workers) )(RRC (VP (VBN exposed)

(NP (-NONE- *) )(PP-CLR (TO to)

(NP (PRP it) ))(ADVP-TMP (NP (QP (RBR more)

(IN than)(CD 30) )(NNS years) )(IN ago) ))))))

Extracted Output :###########VB: exposed#Subj: NP-SBJ#Arguments: NP#DirObj//PP-CLR#PrepObj(to)##########

Figure4: An exampleof extractionfrom thePennTreebank

This work is still underway 19. From the ab-stractlevel of syntacticgeneralization,a TAG willbeautomaticallygenerated.It is interestingto notethat theresultinggrammardoesnot have to closelyreflect the linguistic choicesof the annotateddatafrom which it was extracted(contrary to type Bgrammars).Moreover, from thesameabstractsyn-tactic data,one could also generatea grammarinanotherframework (ex. LFG). Hence,this abstract

19For now, thisprojecthasalreadyyielded,asabyproduct,afreely availableprogramfor extractingverb subcategorizationframes(with syntacticfunctions)from thePennTreebank

48

Page 53: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

level maybeviewedasasyntacticinterlinguawhichcansolve someportability issues20.

6 ConclusionWe have proposeda classification of grammardevelopment strategies and have examined theadvantagesand drawbacks of each of the fourapproaches. We have explained how “grammarreplication” may prove an interesting task tocomparedifferentdevelopmentstrategies,andhavedescribedhow grammar replication is currentlybeing used in the Xtag project at the Universityof Pennsylvania in order to comparehand-craftedgrammars,grammarsgeneratedwith MetaRules,and grammarsgeneratedwith a MetaGrammar.We have reachedthe conclusionthat of the fourgrammardevelopmentstrategiesproposed,themostpromisingone consistsin automaticallyacquiringan abstractlevel of syntacticrepresentation(suchastheMetaGrammar).Futurework will consistinpursuing this automaticacquisition effort on thePennTreebank. In parallel, we are investigatinghow the abstractlevel we acquirecan be usedtogenerateformalismsotherthanTAGs(e.g.LFG).

Aknowledgements:We thank the Xtag group, and more particularly W.SchulerandR. Prasadfor helpful commentson earlierversionsof this work. We alsothankB. Crabbe andB.Gaiffe for their help with the LORIA MetaGrammarcompiler.

ReferencesA. Abeille, M. Candito,andA. Kinyon. 1999. FTAG:

currentstatusandparsingscheme.In Proc.Vextal-99,Venice.

A. Abeille. 1991. Une grammaire lexicaliseed’arbresadjointspour le francais. Ph.D.thesis,Univ. of Paris7.

N. Barrier, S. Barrier, andA. Kinyon. 2000. Lexik : amaintenancetool for FTAG. In Proc.TAG+5, Paris.

T. Becker. 2000. Patternsin metarulesfor TAG. InAbeille Rambow, editor, Tree Adjoining Grammars,CSLI.

M.H. Candito.1996.A principle-basedhierarchicalrep-resentationof LTAGs. In COLING-96, Copenhagen.

J. Carroll, N. Nicolov, O. Shaumyan,M. Smets,andD. Weir. 2000. Engineeringa wide-coveragelexical-izedgrammar. In Proc.TAG+5, Paris.

20The notion of “syntactic interlingua” was usedin otherpapersas an analogy to the terminology usedfor Machinetranslation: “simple” grammarextractionalgorithmscouldbeseenas “transfer approaches”(i.e. low level of abstraction)whereasMetaGrammarextraction could be seenas “interlin-gua” approaches,in thesensethata higherlevel of abstractionis needed(the “lingua” being a syntacticframework suchasTAGs,LFG etc.)

J. Chen.2001. TowardsEfficientStatisticalParsingus-ing LexicalizedGrammaticalInformation. Ph.D.the-sis,Univ. of Delaware.

D. Chiang. 2000. Statistical parsing with anautomatically-extracted TAG. In ACL-00, Hong-Kong.

L. ClementandA. Kinyon. 2001. XLFG: anLFG pars-ing schemefor french. In LFG-01, Hong-Kong.

A. Copestake. 1999. The (new) LKB system.In CSLI,StanfordUniversity.

C. Doran, D. Egedi, B. Hockey, B. Srinivas, andM. Zaidel. 1994. XTAG system-a wide coveragegrammarfor English. In COLING-94, Kyoto.

R. Evans,G. Gazdar, andD. Weir. 2000. Lexical rulesarejust lexical rules. In Abeille Rambow, editor, TreeAdjoiningGrammars, CSLI.

B. Gaiffe, B. Crabbe,andA. Roussanaly. 2002. A newmetagrammarcompiler. In Proc.TAG+6, Venice.

K. Gerdes.2002.DTAG.attemptto generateausefultagfor germanusing a metagrammar. In Proc. TAG+6,Venice.

A.K. Joshi. 1987. An introduction to tree adjoininggrammars. In Mathematicsof language, JohnBen-jaminsPublishingCompany.

A. Kinyon andC. Prolo. 2002. Identifying verb argu-mentsand their syntacticfunction in the PennTree-bank. In LREC-02, LasPalmas.

A. Kinyon. 2000. Hypertags. In COLING-00, Sar-rebrucken.

A. Kinyon. 2002.Featurizinga treeadjoininggrammar.In Proc.TAG+6, Venice.

S.Lehmanandal. 1996.Tsnlp— testsuitesfor naturallanguageprocessing.In Proc. COLING-96, Copen-hagen.

M. Marcus,B. Santorini,andM. Marcinkiewicz. 1993.Building a large annotatedcorpusof English : thepenntreeban.In ComputationalLinguistics, Vol 19.

R. Prasadand A. Sarkar. 2000. Comparingtest-suitebasedevaluation and corpus-basedevaluation of awide-coveragegrammarfor English. In LREC-00,Athens.

C. Prolo. 2002. Generatingthe Xtag englishgrammarusingmetarules.In Proc.COLING-02, Taipei.

J. RogersandK. Vijay-Shanker. 1994. Obtainingtreesfrom their description: an applicationto TAGS. InComputationalIntelligence10:4.

A. SarkarandS. Wintner. 1999. Typing asa meansforvalidatingfeaturestructures.In CLIN-99, Utrecht.

F. Xia. 2001. Automaticgrammargeneration from twoperspectives. Ph.D.thesis,Univ. of Pennsylvania.

XTAG ResearchGroup.2001.A lexicalizedtreeadjoin-ing grammarfor English.TechnicalReportIRCS-01-03, IRCS,Universityof Pennsylvania.

49

Page 54: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

���������;����������K�B�����������������������$�%������� J���.¡����%�����¢�¤£�¥��B�¦�%���¦�¨§�©ª«����������$������¬­�B��¥®�¤�%�����¯�°®±�²´³Bµ·¶¹¸¦º¼»v±$½¿¾º�½.ºÀ;ÁÃÂYÄ|ÅÇÆ�Á´È�Á´É´Â Ê Ë­ÌEÁÃÍ$ÎYÂÁÐÏ�Ñ�ÂYÄ�ÒBÁÔÓ¹ÀÐÆ�ÌÕÏ�Ö× ÌEË�ÁÃØÇÙÚÍÜÛ$Á®Ý­É´Ñ�ÒBÁÃÂÎYÑ�ÙWÈÞ>ß�à|á4â ÝvÁÃã:äWÉ´Íæå6çèÂMÉ|ÍBÊJÁÌÕÉ|ÂYÄ�ä\ÙÚÍ�Á�éëê;É|ì>ÁÃì�Á>í�Å: ÊJÁ�éîÅ:ÁÃÂÄ´Å.éfÊ Ä�Øïð±�ñ¦²´¶¹º$µ�ïóòEô澺$õö Æ�÷�øúù�Í�ÙÚû�ÁÃÂMÈ�Ù\Î�üÁþý�äWÉ|ÙÿÈ�Á��¿É�ÈÊ�É|äÞ|á�� û�ÁÃÍ:Ñ�Á ÌÕÉ|ÂYÍ�Ä�Î×�Þ>â�â>â ÌEäÚÁÃÂYØÔÄ�Í:Î�ç�ÁÃÂY ɴÍBÛÔåEç� É|ÍèÊuÁö É���ÂÙ\ÁÃäbéëý�Á´ÈAí�Ñ�Í�Ù\û$å���Ò.Êuä\ÁÃÂØ Ä�Í:ÎÃé��WÂ

Ôñ¯õ�´²´± ������������������������������ "!��#�%$&�(')!�*+�-,.���������0/1�� �2�3�����!��546 �*�$�/��87"2��9�� �7.�����: �7"�� "!�/�';*�<=�> ?';��/�2@*�����!���AB�C4DA;�@4����FE �����'G�> �7H*�<I�� "'����>��!��C4J3�/5�>�K2@�L�>ABA;��!��C4��>/�<M*@��AN�>/1����A�O�������P���������Q<M*���AR�87)!�����*@3�2@�TSDUCV@W�X"U�Y�ZMU\[�]�^�_J`�a8^:ZMUCa:Y�b!�����!c$��1/�/�d��K�����:���� "!��(7eO=fg������!8b"!����9��!L�>!�3��c*�< /1�� �2�3�����!��546 �*�$�/��87"2��h�� i2@�J��ARAN�(���j���k7.�1�J4J3������87ebl!����: i!����Bmn�1 .E2@3��1��!��C4 o ��*����:��!������D�>���D���������: ?!��87p�� �7j!Q$c*��J,���������AB�� "!��������AB�� "!��1*� ��87eOg������'p��/�/13���!��J��!���!���������3��L�>d���/��1!Q'p*�<e!����/��� �2@3�����!��C4��� "<M*���A;��!���*� h�� �4J/�*@���(7B�� ;!��������hoq��*@������!������8O

r ïð²´± sts ±%² ±�¸Iu ²´ºwv¦õ�±�ñ¦¶¹µ¹¶x�y ³gzµ¹¶¹¸¤½=v¯¶·õ�y¶���{�¸¯³e| µ·ºwu¤½.º} ��K*�< !����K4J�� "!��J��/w��*@�� "!��c�� B/��� �2@3��1��!��C4��>/1/�'�EQA~*@!��F�@��!��87 ��>!�3��J��/c/5�> �2@3���2@�p����*.4-�������1 �2�����!����B �*�!���*@ i*><&2��L�>A~EAN�>�8Ol�Q!��1�K4J*@ARAB*� �/�'h�@4�4-����!��(7�!����>!I�~2@�J��ARAN�(����������M�I�H�@�@���e���w�� "!��: �7"�87eb"��AR*@ �2c*@!������ !����1 �2@�8b(!�*Id��=d�*@!���~�����84J�����~!�*"*@/g*�<= ��>!�3��J��/ /C�� �2�3���2����8��7"���J4J������!��1*� G�> �7!�����7"�84-/5�>�L�>!����@�P7��>!L����*@3��J4J�P$����54��0A~3���!�d��i�1 "!����QE������!��87�d?';�~4J*@AR��3�!����8O�������cAB�8�> ��D!�����!c!����K�L�>AB��AR��!L�>/5�> �2@3���2@�j�M!����I2��L�>A~EAN�>!��C4��>/@<�*@��A;��/�����AB�����c3����87�!�*��: �4J*�7"�I!����97.�84J/C���J��!������/��� �2@3�����!��C4c�� "<M*@��AN�>!���*@ �� �� �7�!����j��3�/����c!�����!c$��1/�/"<����87;����>�����:�8O�i��4J*� ����C7"�:�N!�����!R!������;7"*@3�d�/1�R<M3� �4J!���*� �*�<���2��L�>A~EAN�>�=���I�~7"���J�@7"�@�> ?!J��2���<M��*@At!�������*@�� "!�*�<��"����$�*�<D!�������:3��L�>d���/1��!Q'K*�<�/1�� �2�3�����!��54 6 �*�$�/1�(7"2@�@Og�Q �7"�:�87eb�!����c�L�>AB�7"���"�C4J�p�M!����I2��L�>ABAN�>�L�g4J*@ "!L�>�� ��=d�*@!��B/��� �2@3��1��!��C4l�1 .<M*@�xEAN�>!���*@ ;�� �7H��*�AB���� "<�*@��A;��!��1*� B�@7��>��!��(7�!�*��������84J�F��42@*"��/e������������ �2R�M������/�/�*�$P*��= �*�!L�:b@2��� ����J��!���*� wb"��!L4>O �~�> �7!�*��R�����84J�F��4~�>/�2@*@����!���AR�54�AN��4����1 �����'"OK�Q �*��L7"�:��!�*;d�����:3��L�>d�/��&�> �7j��3���!L�>d�/�� <M*��w7"�F�������� "! 2@*?�>/��8b�$c�D��*�/C7I!�����!/��� �2@3�����!��C4 6 �*�$�/��87"2���A~3���!jd���<M�����p<M��*�A��� "';�����84J�F��4���(�@3��1����AB�: ?!��8b�$����1/��Dd����� �2�b� ���������!�����/������(b"<M*���A;��/�/�'=�J,"E�����������87eO�f�3���!�������AR*@���@b��1!�A~3���!�d��hAR*.7.3�/C����/�'p*@��2?�� .E�����87�$��1!����� �7"�F�e������ "!K/����@�:/1�j*�<D�J,���/��C4J��!I2��L�� "3�/C�>����!Q'B�� *@�J7"����!�*h*>�����I�k!L�('"/�*@�����J��d�/��I�@4:4J�����(O�me�� �2@3��1��!��C4�o ��*@�"E����!������8b $����54���$=�B7.�1��!��� �2�3��������87H<���*@A�o ��*�4J�:�����:�pEI��O �@O�J�e�84J!��1���h4J*�AB��3�!J��!��1*� ���/�����*�4J�87.3������K*� B��!����� �2���*�<��Im�J,������������1*� ��pEh����������*�������d�/��R$&�('i!�*;<M3�/F��/�/c!��������;���JE��3�������AR�� "!��(O

� � ¶·¸�½=v¦¶¹õ�´¶����в´³l�Eº:²"´¶·º:õmn�1 �2@3�����!��C4io ��*�������!��������M*@������AR��/�' o ��*����:��!������J�i$c�����*�����2@�� ��>/1/�';7"���@�:/1*����(7P�� �!����R/C��!��;�@��� �K$���!����1 i!����N¡�oo ���J�@7"��2�A�Oi������¡io �~��!J�� �7¢<�*@�B]�^�_@Z�_8[J_�£�a8bl��'"��!���A~E�>!��C4�*�d��������@��!��1*� ��c*@ B���� "!��� �4J���8be]�^�_J`�a8^:ZMUCa:Y�b@/��� �2�3�����!��C47"�(4J/C���L�>!����@� 6 �*�$�/��87"2��@bI]�^�_�¤�aJ[(ZMU\_@V"Y�bc2@�� ����L�>/1��¥8�>!���*@ ��*� �oq��*@������!������K*�<D��*�AB�K ��>!�3��J��/w/C�> �2@3��>2@�@b�]�^�UCV�[8U `�£1a:Y�b4J��*@���QEQ/1�� �2�3�����!��54j4J*@ ���!��J���� "!��c*@ ;o ��*8¦-�84J!���*� ���*@��o ���� �4-�§E��/������� �7B]�^�_8[Ja:YLY�a:YI!�����!������j�J��<M�84J!��1����4J*@AR��3�!L�>!���*@ ���/����*�4J�87"3������8Oc�i�K���������� "!j�1 ;!����������84J!���*@ �!���� 6 ���� ���/ *�<o ��*@������!�������¨�O������<M*�/1/�*�$��� �2~�J,��>AB��/����� "!���*.7.3�4J���p�� ?!�3���!����@�:/1'�o ��*@�"E����!������K�� �7�!��������j��*�!��� "!��C�>/e*�<�AR*�7"3�/C�����1!Q'"O© ������ �!����;<�*@/�/�*�$��1 �2�f����� �4����J,�������������*@ ��1 ����5�:b���!~�����*@������d�/��D!�*K7"����!��1 �2@3������gb��� ���!��gAB��!J��/C�� �2�3���2��@b�7"�F�e������ "!/C�('@�:���c*@�I�������(4J!��(b���/�/�3���!��L�>!��87Rd"'����§E��@�l!�*;���FEx4��:O��/�����!���*��1�cª���3����

�§E���«K/����QE�¬�����!8b�7"�J<�­�b�!���*@���QE�¬�4����J7�­�b�ª���3����QE�¬ "­=®�FEQd#�D/����xE-¬��>��!8b�7.�J<�­ ¨ !���*@���QE�¬�4����J7�­M¯�ª���3����xE-¬ "­C°��Q±c²�FE�4i³´�B/����xE-¬�����!8b�7"�-<�­ ¨ !���*@���xE-¬�4��>�L7�­ ¯ ª���3����QE�¬ "­ ° � ±c² b«�³�µ@b�¶"·�b�³I¸�b�¶"·�b�³I¶�b�¶.·9®�·���5�n���q�j��!����� �2K*><��J,�������������*@ ��(¹��M�FEx���e���D�j����!(b@�8��4��~AB��A�Ed�����*><&$����C4����1���> ��J,�������������*@ P*><9���C�9�>����*.4-�5�>!��87�!�*��4��>!���2�*@��';�M*@�K[Jº@Z�b@�����Kd��:/1*�$I�:b"�@O 2�Oc¬�����!(b?7.�J<�­�¹��M�FEQd��l�1�c����>���� "!�����!������87�/�����!j*@d�!J���� ��87H������/�'"�� �2~*@�J7"���K���:/5�>!���*@ ��!�*��M�FE��@�:b��> �7;�� �7"�-,.�� �2R��!9$���!���!����p�C7"�� "!��F���:�k»�V¢�C<M*����*@AB�� ��>/� "3�4-/1�(���I�����L�>���8�:¹D���§E�4����1�j*@d�!J���� ��87H������*�4J�C��!QE�� �2��M�FEQd��j!�*�������!p*�<����������~³½¼g¾L¿�·�bw$����:���~`��� �7�À�>���p��*�����!���*@ ����� P���§EQd��:¹e!����R�������K/��� 6 �I!����~��/���AB�� "!K�� ��*@���1!���*@ ~`�!�*~!������:/1�:AB�� "!c�1 R��*�����!���*� GÀ ¯ O�i�~$���/�/D�L�('i!�����!R�M�FEQd����1�p�PÁJº�Y�UC[;Âp_8Ã�a�£j���87"3�4J�87�!�*��!��;Âp_8Ã�a�£�Y�Z�^�UCV@W�bD ��>AB��/�'¢�M�FEQd���bl!����>!;�M�FEx�������~�;`�º@[JÄÅ�Æ�Ç-ÈwÉDÊ�ÇJË�Ì�ͧÎ�ÏMÎeÉ-Ð@ÑcËIÇJÈ�Î�Ò�ÇJÈ�ËIÉJÍ�Ó§ÔMÎ�Ñ=Õ:ÎQÈQÔCÓ§ÇJÐ�ÖJÔ�Î�ÎD×�ØeÙwÚ(Û�ÜQÝ

Þ�ß ÎwàJá�á�ÉJÈQÉ�Ñ(Ó§âJË�ã�É:Ô�Ì�ÈQÎ�Ô�ÎQÐ�ÏMÎ�Ñ É�Ô�Ô\ä@Ê ß Ó�ÐK×1ØDåæ:æ Ü�ÉJÐ@ÑI×�ØeÙwÚ(Û�ÜÏ ß ÇJä@â ß ÉJÐ�Ï�Î�Ê�Î�Ñ�ÎQÐ�ÏMÔ�ÖJÊQÓ§Ï�Î�ÑjÓ�Ð~×1ØeÙgÚ8Û�ÜxÖçâ:Çcè@É�Ê�éIÏMÇ�Éjê æ ÈQÎQÌ?Ç-È�ÏÏ�ÇhÏ ß ÎKë�ì(á�íwî Þ�ï æ ïhðwñeò ígó#ÎQä�ÈQÇJÌ"Î�ÉJÐ~á�ÈQÇJôMÎ�Ê�Ï�Öeã ß ÎQÈQÎ9Ï ß ÎÐ@Ç�Ï\Ó§ÇJÐIÇJÒ�Ñ�Î�Ô�ÊQÈ�Ó�Ì@ÏCÓ§Õ�ÎnËIÎ�Ï�ÉJͧÉJÐ@âJä�É�â�ÎJÖ�Ï�Ç(Ñ�ÉLõ�ö Ô�á�ÈQÇJÌ?ÎxÈ�Ï\Ó§Î�Ô�Ö�ã�ÎQÈQÎÎ�÷�Ì�Í�Ó§ÊxÓ§ÏMÎxͧõ�Ó�Ð�ÏCÈQÇ(Ñ(ä@Ê�Î�Ñ�ÝiÆ�ÇJÈKÌ�ä�è�Í�Ó§Ô ß Î�ÑBã�ÇJÈ�é?Ö ÔMÎ�ÎB×�ØeÙ ñ æ:æ ÜQÖ×�ÙwØ æ ê(ÜQÝ;Æ�ÇJÈ9ÉpÑ(Ó�ø?ÎxÈ�ÎxÐçÏlè�ä@ÏlÈ�ÎxͧÉ�ÏMÎ�Ñ~ÉJÌ�Ì�ÈQÇ�É�Ê ß Ö�Ô�Î�Îp×1ØeØ æ:æ ÜÉJÐ@ÑK×�Ø�ͧÉ�Ú�Ú(Ü�ã ß Ó§Ê ß È�ÎxËIÉJÓ�ÐIù�âJÈQÉJË�ËIÉJÈ�ÇJÈ�Ó§ÎQÐ�ÏMÎ�Ñ�ù�Ýú ð Ð@Ç�Ï ß ÎQÈ�ͧÉLõ:ÎQÈJÖ8Î�÷:Ì�ÈQÎ�Ô�ÔCÓ�Ð@â É�Ô�ÎQËIÉJÐ�ÏCÓ§Ê�ÈQÎQÌ�ÈQÎ�Ô�ÎQÐ�Ï�É�ÏCÓ§Ç-Ð?ÖJÊ�ÉJÐè"Î�É:Ñ�Ñ�Î�Ñ"Ö�è�ä@Ï�Ó§Ï�Ó§ÔeÐ@Ç�Ï�Ñ(Ó§ÔMÊxä@ÔMÔ�Î�ÑcÓ�ÐKÏ ß Ó§Ô�Ì@É-Ì?ÎQÈJÝ

¡@�

Page 55: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

������*�4J�C��!��87R!�*R�M�FEQd���b�!����>!=���������j³G¼g¾L¿k· �����kûK^:^�_@ülUCV@W`�º@UC^JY�bg!����>!~�M�FE�4����1�p�� �º@^:^�_@ülaLÃHÂh_(Ã@a:£��1 �4J*@����*��L��!��� �2!����9AR*�7"��/e��!����� �2;���§EQd��:O�i�97.�1��!��� �2�3������~!���������d������C4 6 �� �7.�=*><�o ��*�������!������pýDþg,"E����!��� �4J�Po ��*@������!������8b�mn�1 ��8�����1!Q'�o ��*@������!������8b�ÿI����*�$��� �2o ��*@������!��1�:�8O;o���4 6 �p������ ¢�M�FEx���K�����p�����84J�F���(7id"'Gþ�,.���xE!��: �4J�ho ��*@������!������8¹�*@�L7.��������/C�>!���*@ ��(b@d"'Bmn�1 ��8�����1!Q'~����*@�.E����!������8¹�ÿj����*�$��� �2p���>�1���8b"d"'Bÿj����*�$��1 �2�o ��*@������!������8OþD��4����C7"�� "!��F���:�BE~�@O 2�O)»�V�E~�����;��!��;������*�4J�C��!��87�����!*�<Io ��*@������!��1�:�8bc�@O 2�O;]�^�_-`.a(^�Z�UCa:Y��M»�V�b����M»�V¢d����� �2�!�������:!�*�<���/�/��� �7)*@ �/�')!���� AR*.7.��/�� Â��M»�Vt�J��!��1�x<M'"�� �2o ��*@������!��1�:�QE��� wOco ��*�������!������K�����j�J,������:�����(7;*@ ~��'?A~d�*@/��$����C4��B�����h[Jº@Z�� YI*��=�C7"�� "!��§�������(ODf���*@A)�������8�(<M!�����$c�I3�����  �>�N� AB��!J�(�@�����C�>d�/��B*@ ��57.�� "!��F�������8b��� �7 � �����AB�:!L�(�@�����C��d�/���*@ �[Jº@Z�� Yk�> �7 � Â�� Y�Oÿ#[Jº@Zl���=�K���:!=*><w/5�>d��:/ �>�@��/�3��c���������8b�*@�8b��� ~���87.3�4J�87R �*>E!L�>!���*@ gb��p���:!9*><D�@��/�3����������j�� N�������"��3�*@�j�J,���AR��/����L�:O=ÿAN�ç,��1Ap3�A)4���!���2@*������p�LÂp[8�c���I�B[Jº�Z=!�*~$����C4��B �*~*�!�������@��/�3��~4��> �d��B�@7�7.�87eO~�����~������3�AR�87�me�J,��C4J*@ ����h�R����!*�<w/��J,��C4��>/��� "!��������8b"�(�@4��;*@ ���d����� �2~�� B�-,.������������*@ ;������*�E4J�C��!��87H!�*B*� ��p*@�KAB*����~Âh[�� Y;¹w4���!��=��3�d���3�AB����4��>!���b��F<4��>!����04���!���O�����~$���*�/��h��'?��!���A 4��> �d��~�"����$=�(7��>�h�RAB*�7"3�/C���I�ç,"E��*@A;��!��C4l��'"��!���A �1 R$����C4��BAB*�7"�:/1�c�����K!����K*@d@¦J�84-!����L�>!QE���Q<M'"�� �2I7"�F�e�����: ?! 6 �� �7.��*><�o ��*@������!��1�:� ° Ogÿ�d������C4 AR*�7"��/Qb��� �� R�M�FE�d���b�$���!��p��!��D������*�4J�C��!��87p����4 6 b"��� �� B���FEx����b(�L�>!����QE������þg,�����!��� �4J�~�� �7�me�� ��8�>����!Q'�o ��*����:��!������(¹w�� �������*�$=�(7AB*�7"�:/�b������� ����§E�4��l�L�>!����Q���:�9��/���*�ÿI����*�$��� �2~o ��*�������!������(Of�3���!�������AR*@���@b�2@���"�� �2��j����!c*�<wo ��*����:��!������(b��jAB*�7"��/�4��> �L�>!����Q<�'~��*@AR�@b�d�3�!� �*�!&�>/1/�*�< !����:A�O o ��*�������!������K4��� ;d���J,������������87��� �7.�����: �7"�� "!�/�'¢!����;*@ �����<���*@A !����;*@!��������8b�� �7i�� ��> ?'i*��L7"�:�8O¢�����;����!~*><I<M�8�>!�3������p<M��*@A $����C4��4��>!8� ���>���Rd�3���/C7i4��� ¢d��RAB*@���~*@��/������h�-,.!��� �7"�(7eb��� �7eb4J*� ����8��3��: ?!�/�'?be!����~2@�J�� "3�/C������!Q'B*�<K[Lº�Z�� Y-bg�� �7;*><&o ��*@�"E����!������I�-,.���������87�*� ;!�����A�b@AR*@���I*@��/����������J�� ��87eO������AR*�7"��/���3�d���!��1!�3�!���*� ���3�/�������/C��!����B!������C7"�: ?!��F�������� AH¨-bpO�O�OTb � A ² bj�8�@4���*@ ���$���!�����!����>����*.4-�5�>!��870�NE� A���OH�Q ��;A�E � A��c$��1!��P� � A��?b ��!���3�d���!���!�3�!����~��*�AB�A~E � A � <M*@� � A � O�þ�O 2�O�be������3�AR�� �2���V�aQW ¨ º�Ã�¤�¯��! #"%$�²������f����� �4��PA�E�ÿ'&)(� P��3� �7"�:��/���'? �2�bD��O 2�O�!����B��!����� �2H`�º�YÁJa�£\£1a:Y8��b�!����KAB*�7"�:/���3�d���!���!�3�!���*@ ~��3�/��I*�d�!L�>�1 ��I�Q¸@�g<M��*@A!����I<�*@/�/�*�$��1 �2~�Lµ(��Oµ@Oj³ ���>��! ¨ ÿ*&+(@ ¯ ° � ±-, bD«�³kµ�b�¶.·�bD³I¸�bD¶.·�bD³I¶�b¶.·9®�·¸�Oj³����>��! ¨ ���2"¯I�@7:¦J°� /.8�x± , b�«@³�µ�b10�·�b�³�¸�b?¶.·�b�³I¶�b0�·kb�³20�b/0�·K®�·

�Q B�� p*@��!���AN�>/@����!�3���!��1*� wb@o ��*@������!������&�>����*.4-�5�>!��87p!�*�!����3 Þ�ß Î�Ô�õ(ÔMÏ�ÎQËiè"ÎQÐ@Î54@Ï�Ô"ÒFÈ�ÇJËPÏ ß Î Ê�ÇJÐ@Ê�ÎQÌ@Ï�Ç-Ò?Ò�É�Ê�Ï�ÇJÈ�Ó76QÓ�Ð@ânÈ�ÎxͧÉ98Ï\Ó§ÇJÐ@Ô ÇJÒwÔMÏ�ÉJÐ@Ñ�ÉJÈQÑjÌ�È�Ç(Ñ(ä@Ê�Ï\Ó§ÇJÐjÈMä�ͧÎ�Ô�Ç-ÒeÐ@ÇLã�Ï\È�É�Ñ(Ó§Ï\Ó§ÇJÐ@ÉJÍ"âJÈQÉJË28

ËIÉJÈQÔ�Ý�ì(Î�ÎcÓ�Ð~Ì@É-È�Ï\Ó§ÊQä�ͧÉJÈ Ï ß Î2:"á�ÔMÏ�É�Ï�ÎQËIÎxÐçÏ�ÔwÇJÒ�; áeì1;PÑ(Ó§Ô�Ô�Ç98ÊQÓ§É:ÏMÎ�Ñ~Ò§ÈQÇJË)Ñ�ÇJË�Ó�Ð@ÉJÐ@Ê�Î�î\ó�È�ä�ͧÎ�ÖwÉ-Ð@ÑBÑ�ÎQÌ"ÎQÐ@Ñ�ÎxÐ@Ê�õ�âJÈQÉJË�ËIÉJÈ×�× Þ Î�Ô�< æ Ü�ÜxÖ�Î�ÉJÈ�ͧõ�Ùwáeì1;¢Ó�Ð�×�á ð ê�=�ÜxÝ Þ�ß Î�ÔMõ(ÔMÏ�ÎQË0Ï\ÈMÓ§Î�Ô=Ï�ÇKÌ�ä@Ô ßÏ ß Ó§Ô�è@É�Ô\Ó§Ê�Ó§Ñ�Î�É Ï�Ç Ó§Ï�Ô�Í�Ó�Ë�Ó§Ï�Ô�ÖJÑ(Ó§Ô�ÇJͧÕ�Ó�Ð@âgÏ ß ä@Ô�Ï ß ÎwÊ�ÇJÐ@Ê�ÎxÌ@Ï?Ç-Ò@Ì�ÈQÇ98Ñ(ä@Ê�Ï\Ó§ÇJÐ�È�ä�ͧÎ�Ô�Ý ð Ð�ÉJÐ@ÉJͧÇ�âDÇ-Ò.ã ß É:ÏwÏ ß Î�Ô�õ(ÔMÏ�ÎQËPÇJÒ�á�È�Ç-Ì?ÎQÈQÏ\Ó§Î�Ô�Ó§ÔÎ�÷:Ì?Î�Ê�Ï�Î�ÑgÏMÇeÎ�÷�Ì�ÈQÎ�Ô�Ô@Ê�ÇJË�Ì@ÉJÈQÎ�ÑeÏ�Ç�Ì�È�Ç(Ñ(ä@Ê�Ï\Ó§ÇJÐeÈ�ä�ͧÎ�Ô�Ö�Ê�ÉJÐ è"Î�Ô�Î�ÎxÐÓ�Ð�ÈQÎ�âJä�ͧÉ-È�Î�÷:Ì�ÈQÎ�Ô�ÔCÓ§ÇJÐ@Ô�É�Ô Ê�Ç-Ë�Ì@ÉJÈ�Î�ÑIÏ�Ç�Ì�ÈQÇ8Ñ(ä@Ê�ÏCÓ§Ç-ÐIâJÈ�É-Ë�ËIÉJÈ�ÔÇJÒ�ÏCõ:Ì?Î ï Ó�Ð9Ï ß Î ñ�ß Ç-ËIÔCé�õ@ö Ô ß Ó§ÎQÈQÉJÈQÊ ß õ:Ý

� Â�� Y�*><=��*�AB�h�jmDb�!�*�2@��!������K$���!����;me�-,.�C4J*� ��� �7;!����AR*.7.��/K��3�d���!���!�3�!���*@ 0��3�/��@bK�����84J�F���¢!����¢$���*@/�������!�*�<AR*.7.��/������8��3������87K!�*I7"���J4J����d��=!����D��!����� �2��w*�<"�J,������:������*� ��*><�!����I�ImlO��i�I4J*@ �4J�� "!��J��!��j�� h!�����<M*�/�/1*�$��� �2��� p!����j�� "E!�3���!����@�K���������� "!J��!���*� �*�<=o ��*@������!��1�:�9�����(4J�F<M'"�1 �2~AB*�7"��/��*�d�!L�>�1 ��87K$���!���*�3�!e!����=AR*.7.��/@��3�d���!���!�3�!��1*� h��3�/��@O © ���@�� !����h7"�F�e������ "! 6 �� �7"�I*><&o ��*@������!��1�:�8be$c�h$���/�/w�� "!�3��1!����@��/�'4����>�L�@4-!������1����!����k4-*@ �7"��!��1*� ��c!����>!�A~3���!cd��j<M3�/§��/1/��87pd"'��AB*�7"��/��� ;*@�L7.���I!�*h�J��!����x<M'B�8��4��N*� ��K!����:A . O� 3�d���3�AR��!���*� I���w!����=d������C4 ����/C��!��1*� K/1�� 6 �� �2cAB*�7"��/��w�> �7o ��*@������!������8O��i�h��/����(�@7"'�7"�J�� ��87i��d�*��@����3�d���3�AB��!���*@ d���!Q$c���� )[Lº�Z�� Y-O ÿI�N� ����*���!����� �7nbI$c���L�('����:����!����>!>@? ����3�d���3�AB�:� >@? ���F< >@? �BA >@? ��OPf�3���!�������AR*@����b2��1���� ;����!�� > �I�� �7 > �~*>< >DC ��'?A~d�*@/��(b"$=�K�L�(';!�����! > ���3�d���3�AB��� > �p�F<w!������������I��d�� ¦-�84J!������I<�3� �4J!��1*� ;d���!Q$c���� > ���� �7 > �H��3�4��P!�����!p�8��4�� >@C1E �� > �K��3�d���3�AR���h��!��4J*�������������*� �7"�� �2 >@C ² �� > ��OFHGJI KML#NPORQ�SUT�VWSYX[Z1\]�S^Z1Q�N�S/Oþg,�����!��� �4J�Ioq��*@������!�������������*�4J�C��!��(7p!�*K��*�AB�=�RE � A%�����84�E�F���;!���������!B*><h���@4 6 �~<���*@A´$����C4�� �� "'¢A~E � A ���~*@d"E!J���� ��87eOg�i�I7"����!��� �2@3������~����� 6 �� �7"�8ý`_D*�4��>d�3�/C����'h����*@�"E����!Q'?b@a� ��54-�1!Q'H����*�������!Q'?b ��3�4J/���3��~o ��*�������!Q'?bDþ�,.��2��� �4J'o ��*@������!Q'"b�þg,�4J/�3����1*� �o ��*�������!Q'?O�����'_D*�4��>d�3�/C����'Bo ��*����:��!Q'"b�������/�/��87Rd"'b/c E A0« >@C ¨@¾�dedfd�¾ >DC ² ®�J�('?�p!�����!p�8�@4�����'"A~d�*@/D�� P!����B���@4 6 ������*�4J�C��!��87¢!�*i�?hgi>@? �c������3�d���3�AB�(7Bd"'R��*@AR�K��'"A~d�*�/��� �b/c E b��> �7�8��4��i��'"A~d�*�/ �1 jb c E ��3�d���3�AR���p��*@AR�h��'"A~d�*�/��� �!��������4 6 *�<���*@AR� ?kg'>D? � O þ�O 2�Oq�M���� �2�/���!�*� h4���!���2@*��������D�����������/�/��87H$���!���!��������K�@��/�3��8�jf����� �4��ml ±=²on «?Ã@a(Zqpw`�_�YJY9p[Jº@^�Ã1p~V�_@X"V/rqrJr�®����R!������@*�4��>d�3�/C����'P<�*@�ts , �M �*�AB�� ���/ "3�4J/��8�>�L�jf����� �4��������L�>�����~����*�3�2@��/�'"b� �*�AB�� ��>/w4��"3� 6 �J��b�>����3�AB�� �2�Âh[�� Y@ýP¬�7"�:!8bI�>��!8bj7"�J<LO�O�O ­xbj¬�7"��!8bj����!(b��� �7eO�O�O ­xb¬�7"��!(b?7.��A�O�O�O ­xb:$����C4��B�����K������*�4J�C��!��87~�� p!����Kme�J,��C4J*@ R!�*�b���������84J!��1����/�'?b�!������J,�������������*@ ��I«�£1a:Y�pe£�º1pg£1a�rJr�®up�«?X"V^pDX"V�aRpÃ�a�Y!rJrJr5®upw«"[Ja!p=[Ja:Y9pq[Ja(ZMZ�a�rJrJr�®"O�����-a� ��C4J��!Q'Bo ��*�������!Q'?b�������/�/��87;d"'v , c E A « >@C ¨@¾�dfded�¾ >@C ² ®�J�('?�l!����>!D!����:���������I �*�!Q$c*���'?A~d�*@/�� �� ~!����j���@4 6 ������*�E4J�C�>!��87�!�*B� ?hgw>@? ��3�d���3�AB�87Rd"'~*@ ����> �7R!����K�L�>AB���'?A~d�*@/��� xv , c E Owþ�O 2�Ogf����� �4��iy V ±c²�n «?Ã�a8Zqp�[Jº�^�ÃurqrJr�®�J,����������l!����>!�!��������I������ �*I!Q$c*k�>��!��54-/1�:�8b�*���!Q$c*97.��AB*� "E��!��L�>!����@�:�9*��9�> �����!��C4J/��h�� �7��B7"�:AB*@ ���!��J��!��1������ ��B�� �����L�����@O��������3�4J/���3��Ko ��*�������!Q'?b�������/�/��87Rd?'skz�c E A0« >DC ¨@¾�dfdfd�¾ >@C ² ®�J�('?�c!�����!��� B�8��4�� ?hg�>@? !������������=*� ����� �7�*@ �/�'~*� ����*@���1!���*@ ;$���!����� ?3�4J/���3��K��'"A~d�*@/�E�������/�/1�(7�{ � E���3�d"E��3�AB�87kd?'K��*@AB�l��'"A~d�*�/��� |skz c E O þ�O 2�Owf����: �4���»�X c En «?[Jº@^�Ã1phÀ(X"º@V�Zqp�V�_@X"V/rJrJr5®��J,����������;!����>!B»�V������L�>�����4��> ;���(�@�K���I�~��3�4J/���3��I�:�1!������K�h4��>�L7"�� ��>/ �M�@O 2�O�U�£�º�}8X� { ZM^�_@U�Y�� ±=² be*��9�;��3��� "!��§�����p�M�@O 2�ORU�£Dºk}(Xw� { ZM_�X@Y�� ±c² ��b*��&�p �*�3� ��M�@O 2�O�U�£wº�}(Xk�Q£�a�Y){�~=a(X"^JY�� ±=² �:O������þg,���2@�: �4J'Noq��*@������!Q'?b�������/�/1�(7Bd"'>@�2� c E « > ¨�¾�dfdfd�¾ > ² ®�LÆ�Ç-È�É ËIÇ-È�Î�Ò1ÇJÈ�ËIÉJÍ�ÉJÐ@ÑcÊ�ÇJË�Ì�ͧÎ�ÏMÎ"Ì�ÈQÎ�Ô�ÎQÐ�Ï�É�ÏCÓ§Ç-Ð�Ö�Ô�Î�ÎgØeÙ#8CÚ8Û(Ý

¡�µ

Page 56: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

�L�('"�p�M����AB�:A~d�����!�����! � � Y�������/�/D����!���*�< � �j!����>!I�F<=�� !���������4 6 *�<�� ?�g�>@? !������������;�1 �4J/�3�7"�87��¢����!;*�<��'"A~d�*@/�� > �I��3�d���3�AB�87�d"' >@� !��������cA~3���!wd��&�>/1��*���*�AB�>`� �� �4J/�3�7"�87p�� p!����c����4 6 ��3�4��~!�����! >`� ��� ��3�d���3�AB�(7pd?'> ���w¨�ODþ�O 2�O f����� �4���«1�QV^pD[��J® � ±c² «�«?Ã@a(Z\®upw«"[Jº@^�Ã>®UrJrqr�®?b$��������B¬ wb�4L­���!J�� �7.�c<M*@�I4J*�ABAR*@ p �*@3� ��8b��-,.���������9!�����!4J*�ABAB*� h �*@3� ������8��3������h�~7.��!�����AB�� �����*��&�~4:���L7.�1 ���/QO������þg,�4J/�3����1*� �o ��*�������!Q'?b�������/�/��87;d"'>@�B�� c E « > ¨�¾�dfdfd�¾ > ² ®�L�('"� !�����! �F<��� �!��������@4 6 *><�� ?�g|>@? !������������ �� �4J/�3�7"�(7�R����!K*><&��'?A~d�*@/�� > �;��3�d���3�AR�87�d"' >@� be!����� �!��������B��� �*�!9�� �4-/13�7"�87��R���:! >`� ��3�4���!�����! >`� ���I��3�d���3�AR�87Hd?'> ���w¨�O�þ�O 2�Owf����� �4��)«��xÀ8X"º@V�ZJp"`�£ �J® �� ±=² «@«��xº@^�Zqp�U\V�Ã��J®�®�J,����������j!����>!�!�������3��� "!��§������ZM_�X�Yk4��> � �*�!I4J*"�J,��1��!�$���!���� ;�� �7"�-�� ���!�������!��C4J/��K�� N�~»�V;�����L�>���@OFHGeF ��N�T@SU��Z1N�Q9��X'Z1\]�SUZ1Q�N�S�Ome�� ��(������!Q'Bo ��*@������!������K�J,����������K*@�L7.���I����/C��!���*� ��8O ÿ#mn�1 .E�8�>����!Q'Bo ��*����:��!Q'B���D�����:/1/��87Rd"' >DC���� c E >@C ¨udfdfd >@C ² Ow�Q!�L�('"�D!�����!l�§<e�� �� ?jg�>@? !��������j���=�K��'"A~d�*@/���3�d���3�AB�(7d"' >@Cu� �� �7��;��'"A~d�*�/w��3�d���3�AR�87�d"' >@C ���w¨@bg!����p<M*@�xEAB�:�=�����84J�87.����!����9/C�>!�!��:�8Oþ�O 2�O�ý �� �f����� �4��PÂ��M»�V�� Y�b��~��3��� "!��§������ZM_�X"ZP��aRpJ�QY��k�����JE4J�87.������/�/j*@!������i[Lº�Z�� Y-bc$����C4������B�-,.�����������87 d?'0À(X"º@V�Z� ±=² V^p=Ã@a(Zqpe`._>YJY�pc[Lº�^�Ã�rJrJrFHGe� �BZ1Z1\/�xN�T@��X[Z1\]�S^Z1Q�N�S/O�����~d��>���C4p��*�/��h*><&ÿj����*�$��� �2�o ��*@������!������h����!�*;�����84J�F<M'!����~2@�L�>����EI�QO �@O~!����~����!p*�<�ÿI����*�$��� �2;���������IEK!����>!����!����Bd���4 6 d�*@ ���<M��*�A�$����C4��¢!����R����A;�� "!��C4p���������:���� "!L�çE!���*� ~�1�cd�3���/C7eOgÿI ;������*�$��� �2����>�1����ûK^��D�����K���>�1�2�F`�pqÀR�?b$���������`N�> �7;À9�>���j��*�����!���*@ ����� p!����jAB*�7"��/"��!����� �2�b?�> �7$����C4���4:�� pd��c3� �7"�:����!�*"*.7~�>�@�@!���� � �� p��*�����!���*� K`N���xE��*�$���!�*h!���� � �1 R��*�����!���*� �À1��Owÿj �ûK^����=!��?3��9�� ;���J4d���!Q$=�:�� �!Q$=* � � Y�O�ûK^/� Y��>���R�J,������������87¢d"'i������*�$��� �2<M*���A~3�/C����be$����C4��wb �� i!��������p����AB��/1�:��!j<M*@��A~3�/C��!��1*� wbe�����������/�/��87 >@C � � c E >@C ��Ow�Q!D���=�>/���*K��*������1d�/1�c!�*�������/�/�7"���xE¦J*� �4J!������B������*�$��� �2�bg�J,������:������ �2�!�����!p��*@AR� >@C �>����*�$��!�*R����!����:� >@C ��*��I!�* >@C ��O��l'��~2��� ����J��/D4J*@ "���� "!���*@ wb�� "3�4J/���3��B{ � ������*�$��h!�*�����AB����/F<LO © �� ����J��/I4J*� �7"��!���*� ��/���AB��!e!����j�J,������:���������I��*�$=���c*�<eÿI����*�$��� �2K<�*@��A~3�/5�>�@b����xE��3����� �2�b"��AR*@ �2K*�!��������8b"!�����!c!����j������3�/�!��� �2p2@�J�����~A~3���!d��h4J*@ � ��84J!��87ebe�� �7nb"$��1!��;!������-,�4-����!���*� �*�< !��������Jª��J,"E���@��������*�$��� �2p*�< { � b��@4J'�4J/��C4�Oþ�O 2�O�bB�>AB*� �2½f����: �4�� ÿj����*�$��� �2�<�*@��A~3�/5�>�@bp!����:���0���À(X?º�V�Z � ±c²�{ � bi$����:����{ � ���0�t�@�>���C��d�/��%*@ !���� ��3�4-/1�:3����� �7%$����C4��t�J,�����������!�����!P!���� ��3��> ?!��FE�����iZ�_@X�YG�>����*�$��B!�*��� "'���3�4-/1�:3����;��  �P»�V������L�����@ý³ ��ZM_�X@Y ¨ {�Z�^�_@U5Y ¯ � ±c² pK«@³ ¡�p-¢�·'p9³'¢�p-¢@·K®@·'pK³ ��ZM_�X@Y ¨£1a:Y�¯-{(W@º�^�_@V"YQ°R��±=²#pl«@³ ¡�p�£�·'pD³'¢up�£�·*pl³*£�p�£@·K®"O¤ ¥�¦ �¦µ¹³B²|¶¹¸�½#�¦²´³D�Eº$²"´¶¹º$õ��$c*��J,����:����AB�� "!��h���(���Bd����� �4��>�������87¢*�3�!��� P!����B�J,"E��/�*@�J��!��1*� �*�<�þg,��1��!��� �4J�N�> �7�mn�� ��8������!Q'N����*@������!������8O��Q !����c������!��-,.�������1AR�� "!8b@mn�� �2@3�����!��C4=oq��*@������!������c$=������3����(7!�*�7"�����1����!�����/1�� �2�3�����!��54H7���!J�i��!���3�4-!�3�������3����87�d"' �4��"3� 6 �:�P�� �7)���Io �J,�!��J�@4J!�*@�i<�*@�iow*���!�3�2@3��:���0��������P�¨§)©����@����O´�Q #!����P���84J*@ �7#�-,.�������1AR�� "!8bpme�� �2@3��1��!��C4

ª«

¬­

ª«

¬­

ª«

¬­

ª«

¬­

ª«

¬­

® ® ®

¯ ¯�° ± ±±³²´fµ�¶-·f¸�¹�· º^»1¼

Å ú 3 �

Å`½U¾�¿�À^Á »1ÂxÃeÄ1Å5Æ ¸�Ç�È É�µ5·7È Å Ç�µ5Ç9ÊDË Å ´fÌ�Í Å�Î Å5Ï È Ð�µ Î µ5Ç9µ Î Ñ�¼ È ¼ Òú@½u¿ » ¾ à Á ¸�Ð Å Ë`Ì Å5¼ È ·ÓÈ Å Ç Å�Ô · Å5Æ ¸�Ç ¼ µ5Ç9Ê@Ì�µ5´f·ÓÈ µ Î Ê�È ¼ µ5Ë#Õ�È Ï5Ö µ5·ÓÈ Å Ç Ò3@½u×^Ø Ã Ù Í Ö Ç Æ È Ç Ï�Ò�@½ º^» ¸�¹�·e´fµ5Ð�·ÓÈ Å Çfg��2@3����~µ@ý �����K����*.4-�������1 �2N4������� �<M*@�I�joP�-,.!��L��4J!���*@

o ��*@������!������h$c�����B3����87i!�*���!���3�4J!�3����;/��J,��C4���/l�� "!��������p�� �> |§jo ��© EQ��!Q'"/��I2��L��ARAN�(�l�M���:���P§+�=��Ú@�c�� �7R�5§I�>2?�@�����:O�Q �d�*@!���4��>�����(bw!����hd������C4K�C7"�(�~���I!����h�J��AR�@ý�������*�4J�C��!��!�*I�8��4��B4:��!��:2@*@��'h7"�84-/5�>���87�<�*@���j2��1���� �AR*.7.��/"!�����4J*@A�Ed��� ��>!�*����C��/"�� "<M*���AN�>!���*@ ��>!�!L��4����87p!�*�!������=4��>!���2�*@��'h�� ~�4J�:��!L�>�� N2��L�>ABAN�>!���4���/�4-*@ "!��J,�!(O�i�I7"�:�L4J���1d��K�������9!��������9!Q$c*p�J,���������AB�� "!��c�� ~AB*����I7"�JE!J����/���HGJI Û¨N�ZuORQ-S^L#]�S^Z1N�Ü�S^T�Q�HGJI�GJI Ý'\T`Q�S^L#Qÿ��� ��I2��L���� ��(7~7"���L4-������!���*� B*�<w!����Iog*@��!�3�2�3��������joi���>�d����� 0�@4�4-*@AB��/1�������87¢$���!�� mn�1 �2@3�����!��C4Noq��*@������!��������> �7$c�;$&�> ?!��87¢!�*�3����H!������~/��� �2@3�����!��C4B7"�:�L4J���1��!���*@ ��� ¢*@�QE7"�:�9!�*;�J,�!��L��4J!~�Io �K<���*@A�og*���!�3�2@3������B��3� � ��� �2R!��J,�!��(O�Q ���������!I��!����gb�!������� ���3�!�!��J,�!I����!�* 6 �: ���¥��87��> �7;AB*��QE����*@/�*@2��54:��/�/�'��� ��>/�'?¥:�87�� � � }'Þ o�§ ��ÿ9����Ú@����O¢������ wb!����h!�* 6 �� ���¥��87i�> �7�AR*@������*@/�*�2@�C4���/�/F'B�� ��>/1'"¥��(7�!��J,�!���������JEQ����*�4J�������87nbe��/���AB�� ���!��� �2I���>��!��5�>/�/1'p��*�AB�K��A~d���2@3��1!Q'�> �7p2@��*@3����� �2p*@�c3� �2���*�3����� �2���*@AB�j!�* 6 �� ��c�������"�1*�3���/�'7"�:/1��AR�1!��87H����o � ��Og������ i!����~!��J,�!����h4��"3� 6 �87i�� �7;��E ��>/�/1'"b8�jo ���Q7"�J�� ��(7B��� ����2@3�/C�>�D�J,������:������*� ��=*><�4��"3� 6 �L��>�����J,�!��L�@4-!��87eOKfg��2@3���� µK��3�ABA;�����1¥:���=!����p����*�4J�������� �24����>�1 �<M*@�I�jo¢�J,�!��L��4J!���*� wO} 3��~4��?3� 6 ���;�Q4���/�/��87 ÿ9f �h4-*@ ����1��!��~�� P�H�@����'i����AB��/���>/12�*@���1!���A%�M���:�K�P�¨§)©��@�����g$����C4��h3������D/��� �2�3�����!��C4 ��!���3�4�E!�3���������4��>/�/1�(7�£1aJºu}(a:Yx�j������*�4J�C��!��87;!�*~�(�@4���!�* 6 �� ;*�<D!����!��J,�!k�> �7H!��������K!�*�4-*@ �4��>!��� ���!��~!��������~��!���3�4J!�3������p<���*@A/��J<M!l!�*p����2@�"!c3� "!���/�!����I�: �7R*�<w!����9!��J,�!8ODþD��4��N4J*� �4���!��JE ��>!���*@ ��� "!���*�7"3�4J���h4J*� ���!��L�>�1 "!���<M*��K!����R ��J,�!h4J*� �4���!��JE ��>!���*@ B�> �7eb�7"3����� �2p���������� �2�b"������!c*�< !������>A~d���2@3���!Q'K�����*@/��@�87R���I�p���57.�9�-���84-!9$����� �4J*@ �4:��!��: ���!��1*� R<�����/��8O� *~��/�/13���!��J��!��9�� "!�3��1!����@��/�'~��*�$0*�3��94��"3� 6 ���I$c*@� 6 �(be���QE��3�AB�p$=�~$��� "!K!�*��� ��>/1'"¥��~!����p<M*�/1/�*�$��� �2~��!����� �2H$��1!��!����j<M*@/�/�*�$��� �2K/1�(�(�@���(OûkY9Ã@º�V/ß[(º�Y�1à#á�apÃ@º@V�[La:Y����S^�^â-I ��������/��8�(< �1�I�>����*.4-�5�>!��87�!�*~ûKYã �����9/���ARAN�K������*�4J�C��!��87R!�*BûkY=���9_ã �����k4:��!��:2@*@��'B���I�p7"�J�� ���!��h����!��C4J/��ã ������AB*�7"��/�$��������K!����1�c4���!���2@*���'B�������8�>���c���= �*�A~E�� ��>/�4��"3� 6

¡@¸

Page 57: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ã ��������4��>!���2�*@��'H ����@�:����!J����!��h�RAB*�7"��/ *><= �*@AR�� ���/4��"3� 6

ã �������B4��>!���2�*@��'P ��������B�: �7"�N��AR*�7"��/�*�<K �*@AR�� ���/4��"3� 6

ã �����K����!�*�<�4��>!���2�*@���1�:��!����>!I4��� �<M*@/�/�*�$��D!����1��4��>!��JE2�*@��'R�� B!����1��AR*�7"��/e4J*@ "!L�>�� ��� �*@3�

� * 6 �� +Ã�º@V/ß[(º�Y�������A~d��12�3�*@3��~��/�3��L�>/§EQ �*�3� ½�> �7��@����d��7��� �4J�:���> �7�7��� �4J�(���� �7H���(���~!����p<M*�/�/1*�$��� �2~!Q$c*����xE��*�4J�C�>!��87;/��8�(�����8O��SU��â*Fã �����K/���ARAN�K������*�4J�C��!��87;!�*;Ã@º�V/ß[(º�Y�����Ã@º@V�ß[8ºã ������4��>!���2�*@��'N���c �*@3� ã �����jAB*�7"��/�$��������j!�������4���!���2@*���'B�>�����(�����c���D �*�A~E�� ��>/w4��"3� 6

ã �������i4���!���2@*���')4��� )��!J����!P��AR*.7.��/�*�<; �*@AR�� ���/4��"3� 6

ã ��������4��>!���2�*@��'���/�$&�('"�j�� �7"�h��AB*�7"��/ *><= �*@AR�� ���/4��"3� 6

ã �����K����!�*�<�4��>!���2�*@���1�:��!����>!I4��� �<M*@/�/�*�$��D!����1��4��>!��JE2�*@��'R�� B!����1��AR*�7"��/��1���:AB��!Q'

�� �7��SU��â*�ã �����K/���ARAN�K������*�4J�C��!��87;!�*;Ã@º�V/ß[(º�Y�����Ã@º@V�ß[8º�^ã ������4��>!���2�*@��'N���c�@����dã �����lAB*�7"��/�$��������l!�������4:��!��:2@*@��'��������8����� �1�g�@����d���/4��"3� 6

ã �������D4��>!���2�*@��'~4��> ~��!L�>��!=�jAR*.7.��/@*><�������d��>/�4��"3� 6ã �������I4��>!���2�*@��'N4��> N�: �7���AB*�7"��/e*><��@�:��d��>/�4��"3� 6ã �����p���:!9*><=4���!���2@*���������!�����!K4��� ;<�*@/�/�*�$ !������94��>!��JE2�*@��'R�� B!����1��AR*�7"��/e4J*@ "!L�>�� ��I4J/���!��C4j����*� �*@3� ��8O

ÿ�<�!����p!����;4J*@ �4���!��� ���!���*@ i*�<I!����B/��8�ç<hµB������*�4J�C��!��87i!�*ûKY�b�!����h*� �/�'N��*@������d���/���!Q'B���K!�*N4-*@ �4��>!��� ���!��~/��8�ç<&¸~d��JE4��>3�����!����KAB*�7"��/���!����� �2p*@ ;!����9����2@�"!=*><qûkY=4��� � �*@!�d��4J/�*@���87���/1�(�(<Iµ� ��:�@���j�� �7"���p �*�AB�� ���/e4��"3� 6 �I�> �7R/1�(�(<¶p���= �*@!I�p��*@������d�/��9��3�4�4J�:����*��K*�<D/��8�(<�µ@O�����D����*�4J�����D*�<�4��"3� 6 �� �2I���w���87"3�4-�87�!�*I�����Q<�*@��A#�>/�/@!������*@������d�/���4J*� �4��>!��� ��>!���*@ ���*�<=/��8�(�����j<M��*�AT/��J<M!K!�*R����2@�"!8b�8��4���4J*@ �4:��!��: ���!��1*� ;d����� �2~������!����C4J!��87�d"'R!������������"��*@3��4J*� �4���!��� ��>!���*@ wO} 3��N4��"3� 6 ���;$&�>�~3����(7�!�*i����*�4J������og*@��!�3�2�3�������!��-,.!�� �7�$���� ���@�>/13���!��(7�*� h!����=!J��� 6 *�<w�IoH�J,�!��J�@4J!��1*� ~$���!��!����9������3�/1!���*�<�ÚuÚuä%�����84J������*� ��> �7�Ú�µ@b�¡uä%���84���/�/e*� B!�����Io 7"��!��(4J!���*@ �å¢����*¢�J,��@4J!;A;��!L4���$����B���8��3������87�d�3�!!����h�Io����8�@7�7.��!��84-!��87H�� N!����p���J<M�:���� �4-�B4J*@����3��I�����J,"E!��J�@4J!��87��æ(ì(Î�ÎD×�ÙwÉ:â�Ú�Ú8Ü.Ò1ÇJÈeËIÇ-È�Î Ñ�Î�Ï�ÉJÓ�ͧÔ�Ý

�HGJI�GeF ��S^�Uç�S/O-��Téèj��S^�^â[X+�/Q�Q�S^Z1T�Oÿ0/��8�ç<����I!��?3��I�p��!���3�4J!�3����~*><�!����K<M*@/�/�*�$��� �2K<M*���AT���i����������:���� "!K��!&�>�I�~o ��*�/1*�2h�����87"�C4��>!��8�:Oê S^�^â�ë�ìíÛ+î��2î¨Ý'�/Quî�ïð\�è@ñ!è�îDòôó³Q��/Q�îDK'ó#Q��/Q�îDÛ#\ ê�ê�õ G���������@ýã�öx÷ � ö _@^�à ÷ _�^�Â2�I���D!����K!�* 6 �� �<M*@3� �7p�1 R!����j!��J,�!!�*��> ���/�'"¥��

ã So��Sga8ÂpÂpº9�9����!����h4J*������:����*@ �7"�� �2~/���ARAN�ãùø º�ZP��aQW�_@^ �I����!����k4-*@����������*@ �7"�� �2N4��>!���2�*@��'ã ��_(Ã1úJÃ|����_8Ã�a�£�úLÃ@a(V�Z�U ûDa(^����57.�� "!��F�����D!�����AB*�7"�:/@�� $����54��R!������I4��>!���2�*@��'�4��> ��������8���

ãýü � Z�º@Z�� ü a�W�UCV � Z�º@Z�X@Y������R!������� ?!���2@�:����bhµH*@�;¸AR�8�� ��1 �2~���������84-!����@��/�'¢!����>!9!������h4��>!���2�*@��'�V�a�}8a(^�bº>£1üDº Y�*><=Y�_@Âpa(ZMUCÂpa:YD��!L�>��!���!����KAB*�7"�:/��C7"�� "!��§���87d"'���_(Ã@a:£7úLÃ�a8V�ZMU ûla8^

ãýþ � ZMº�Z�� þ V�à � ZMº�ZMX�Y��D���w!����=�� "!���2����D��b"µD*���¸�AB�8�> "E�� �2����������84J!��1����/�'0!����>!�!����1��4���!���2@*���'tV�a�}8a(^�b~º>£7�ülº YK*@�RY-_�Âha(Z�UCÂha:YK�� �7"�p!����;AR*.7.��/D�� �7"�� "!��§���87d"'���_(Ã@a:£7úLÃ�a8V�ZMU ûla8^

ãý÷ _�£\£�� ÷ _>£C£1_�üDUCV@W>Y��w���w!����D����!g*�<�4��>!���2�*@�������g!����>!�4��> <�*@/�/�*�$�!����~4���!���2@*���' ø º@Zg�� N!�����AR*.7.��/w�C7"�� "!��§���87d"'B��_8ÃWúLÃ��M�����K��AR��!Q'����:!�$����� þ V�à � Z�º@Z�X�Y=���Iµ(�

�i�K4���/�/w�~SgaLº�ÿ�]�º@Z�ZMa(^�V~�p/��8�(< ��!���3�4J!�3����p�� ;$����C4��R!����������!I�>��2@3�AB�� "!j�M!�����$=*��L7�<M*@��AN�l�1�c �*�!��1 ���!L�> "!��C��!��87eO} 3��c����*�d�/���A)�������9���D!�*�7.�87"3�4J��b"<M��*@A)!����Ko ��*����:��!������(b�>/1/"!����ISwaJº�}8a:Yq]�º@Z�ZMa(^�V"Yw!�����!c������ ��84J�����L����'~!�*��� ���/�'"¥��*� ��K!��J,�!8O�HGJI�Ge� �ôS ê �/Q�N�\�T�O���S^Q!�-S^S^T VW�/Q�SU��\�Z1N�S/O

�/]@]�S^�/Z1N�T@�kN�Tý����N�ç�S^TiÜ�\�è`S ê ORQ�Z1N�T@�© ������ �!����i�@*�4���d�3�/C����'ðb *><���*@AB�HAB*�7"��/K�C7"�� "!��F���:�>@? bc��!p�1�p��*@������d�/��@bc3����� �2iþg,�����!��: �4J��o ��*@������!������N�> �7�> �7�me�� ��(������!Q'�o ��*@������!��1�:��!�*�7"�J�� ��h!�����<M*@/�/�*�$��� �2h���JE/C��!���*@ ��D�� _�,t_���Oº��> �7�Á�d����� �2~��/���AR�� "!���*><%b~O¼������ ���@µ�� ºc�����84-�87"�@µpÁ �F<��� h�� "' ?�g >@? 4J*@ "!J���� ��� �2� �/T�è Á(bwº��>/1$��('"�=�����84J�(7"���~Á� ����� ��� º~*@�J7"���;Á��F<&!��������N���h�>!�/��8����!h� ? go>D? �� $����C4��;��!��1����*@������d�/���!�*~�L�('B!�����!pºK�����(4J�87"���RÁI*���!����>!Ác�����84J�87"�:�Bº@O� ��������� ºl�J,���2@�IÁe�F<"<M*��e�8�@4�� ?�g*>@? $��������9ºc�������8�����8bÁ���/���*B�������8�����8O� �����z�� ºK�-,�4-/13iÁc�F<�!��������K���� �* ?hg�>@? $���!���º �/T�èÁ(O�Q!D���=�>/���*���*@������d�/��I!�*h7"�J�� ��I!Q$c*���3�d�����!���*�<�b~b > � �> �7> µ@O > � 4-*@ ����1��!��c*�<w!����I��/���AR�� "!��D*><�_�!�����!=�>���9�>/�$&�('"��>/1*� ��K�� ��pAR*�7"��/e��!����� �2;�� �7R���97"�-�� ��87;!�����<M*@/�/�*�$��� �2$��('eý�Lî5Ð&Ï ß Î�Ò1ÇJÍ�ͧÇJã�Ó�Ð@â Ô�Î�Ê�Ï\Ó§ÇJÐ�Ö�ã�ÎeËIÉ-é8ÎeÏCã�Ç�Ô\Ó�Ë�Ì�Í�Ó 4@Ê�É:ÏCÓ§ÇJÐ�Ô��çÏ ß ÎÐ@Ç�Ï\Ó§ÇJÐ=Ç-Ò?Ô\ä�Ë=è@Ô\ä�Ë�Ì@ÏCÓ§Ç-Ð�è"Î�ÏCã�Î�ÎxÐ&Ê�É�Ï�Î�â�ÇJÈ�Ó§Î�Ô�Ó�Ð�Ð@Ç�Ï�Ï�ÉJé8ÎQÐcÓ�Ð�ÏMÇ

É�Ê�Ê�ÇJä�Ð�ÏwÉ-Ð@Ñ�ã�ÎjÑ�ÇjÐ@Ç�ÏDÊ�Ç-Ð@ÔCÓ§Ñ�ÎxÈwËIÇ(Ñ�ÎQÍ§Ô ã�Ó§Ï ß Ó�ÐpËIÇ(Ñ�ÎQͧÔ�Ö�è�ä@ÏÏ ß Îwâ�ÎxÐ@ÎQÈQÉJÍ�Ó§Ñ�Î�Éqé8Î�ÎxÌ@ÔeÏ ß ÎgÔMÉJËIÎ

¡@¶

Page 58: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

> � A « �! b#"%$'& b(� ��)��ze� � ¾)&��-®> µK����!����h4-*@AB��/1�:AB�� "!L�>��'�*>< > � �� �bf�*@�D�8��4��N4��>!���2�*@��';º�*><�b�b���!����=�>/���*���*@������d�/���!�*h7.�J�� ��!����9����!+*-,/.��>�=!����K����!�*><q��/�/e4���!���2@*���������!�����!c��*������1d�/1'<M*�/1/�*�$0ºK�� N�>!�/1�(����!�*� ��KAB*�7"��/e��!����1 �2�O§I�(�"�� �2�!��������N����/C��!��1*� ��~�� �7 !��������;����!��8bc*@ ��;4��� �7"�JE�� ��H!�������3�d�����!���*><Bb�!����>!B��/�$��('?�(bc��*@AR��!���AR�����> �7 ��������c��!L�>��!��KAB*�7"�:/���!����� �2~�� �7p!����I��3�d����:!���*�<�b�!�����!��/�$��('"�8b���*@AB�:!���AB���B�> �7� ��:�@���;�� �70� AB*�7"��/j��!����� �2�b$����C4��h��� �����84J������/�'p$�����! ���D ��:�87"�87p!�*k7.�J�� ��j!����c/��8�(�@���!�*�2@��!������j$��1!��0*-,/."O�i� 4��>/�/1�(7)!����:���0��3�d�����!��21 > �MÿI/�$��('"�P��!J����!L��b >@>� � *@AB�:!���AB���j��!J����!J��b�s > �Q���������p��!L�>��!J��b-1�3 �Mÿj/1$��('"��� �7���b > 3 � � *@AR��!���AR���D�� �7����� �7 s43 �������@�:�I�� �7������!���!���������7"�J�� ���!��1*� ����> �704J*@ ����C7"�����1 �2�!����i����!�*�<o ��*@������!��1�:�B!�����!~7"�J�� ��;!����BAR*.7.��/����C7"�� "!��F���87 d"'�Â-�� Â%$=�p4��� ;!����� �4J*� ���!���3�4J!p�p����!j*�<D/��8�(<D���>!�!��:�� ��K!����<M*�/1/�*�$��� �2K$��('�ý

ã �����R������!;����2�3�AB�: ?!��1�B�H�@�����C��d�/�����!����>!~$���/1/cd��!����� P�� ���!L�> ?!��C��!��p$���!��P�R/��� �2@3�����!��C4K<�*@��A �����:���� "!�� B!�����!��J,�!L�

ã ���������(4J*@ �7�����2@3�AR�� "!B*><k!�����/1�(�(<h�����87"�C4��>!��i����� ���!L�� "!��5�>!��87R!�*��> N�:/1�:AB�� "!c*�<2b

ã �����p!����1�J7��>��2�3�AB�� "!�*><=!����p/��8�(<c�����87.�54:��!��~���j�1 .E��!L�� "!��5�>!��87R!�*~A~E � A�O

ã ������<M*@3���!��;����2@3�AR�� "!D*�<w!����I/��8�ç<w�����87.�54:��!��K���c�1 .E��!L�� "!��5�>!��87H!�*�µ@be¸�be�R�@4�4J*��L7"�� �2;!�*B!�����<���4J!I!�����!!���������/���AB�� "!c����AB��A~d����D*><51 > b >@> *@�)s > O

ã ��������<M!��P����2�3�AB�: ?!�*�<�!����R/1�(�(<I�����87"�C4���!��;�1���1 .E��!L�� "!��5�>!��87p!�*~µ@b@¸j*@�D�K�@4:4J*@�J7"�� �2�!�*K!����c<��@4-!�!�����!!���������/���AB�� "!c����AB��A~d����D*><51�3~b > 3)*�<�s43~O

ã �����%���F,�!����>��2@3�AB�� "!�4J*@����������*� �7"�0!�*t!����%����!*-,76 ��8 d����� �2�[Lº�Ze!����h4���!���2@*���';!�����!I�������������: ?!K�� !��������(4J*@ �7H����2�3�AB�: ?!(O

�HGeF ó³S/VW\T�è�S^L#]�S^Z1N�Ü�S^T�Q�Q ~!����1�l���84J*� �7R�J,����:����AB�� "!(b�$=�j$&�> "!D!�*�3����I!����Io ��*@�"E����!������P7"�J�� ��87%<M*���!����� �*@AB�� ��>/h4��"3� 6 �� #*��L7"���i!�*4J*� ���!���3�4J!¢/��J,��54:��/~�� "!��������P!�����!P4��> )�� ���d�/1��!�*#�� ��çE/�'"¥��h �*@AB�� ��>/w4��?3� 6 ���� G�> �§Io ��© E���!Q'"/1�R�M�������5© � �10"��� �7;� � � �@�����jO �����[§jo ��© 2@�L�>ABA;���w$����c!����� ;�1AR��/��JEAB�: ?!��87K�� �ÿKmwþ���ÿ�!�!�����d�3�!���me*�2@�C4Dþ �2@�� ��@b@7.���@��/�*������(7d"'k�IO�©��>������ "!����p�� �7 © O�og�� � ���O } �/�'�!����h��'" ?!J�@4J!��C4���>��!�*�< !�����/1�-,.�C4��>/e�� "!����1�:�I����!L� 6 �� ;�� "!�*~�@4�4J*�3� "!8O�i�h7"�(4J�C7"�87H!����>!�<M*��I*@3��I2@�J��ARAN���c�~ �*�AB�� ���/e4��"3� 6���>��!�*�d��I�I�J��!�3��L�>!��87p���12� h$���!����j �*�AB�� ���/����8��7eOw������� ���/�'"�����D<Q����/��=�F<Lý

ã ��*B�� ��>/1'"���1�c��������*�7"3�4J�87ã ÿ%/1�� �2�3�����!��54j���12� ����I*�d�!L�>�1 ��87;d�3�!j��!��1�j �*@!j�L�>!�3"E�J��!��(7

�HGeFHGJI ì:9@�/QB�-S49@�^çS Q�\wVW\T�ORNPè`S^Z�i� ���(����!�*I!J� 6 � �� "!�*I��4�4J*@3� ?!g!����l��!���3�4J!�3��L�>!���*@ p*�<�/��� "E2�3�����!��C4�����2@ ���!�����!2§Io ��© <M*@��AN��/�����A%��!��1��3�/C��!��:�8O �����>!���8ýñ9T Q 9@SýQR�#]�S;9DN�S^Z1�/ZuV 9�� ÿT/��� �2@3�����!��C4p����2@ ¢�����p�� !����I���>!�� ��< � � þD� ý ��< ��ý�m } ©Ký/©cÿD��ý^§Iþ ÿ'&)�C<���*@A �*�$�*@ K!����=$���*@/�� ����!�������7"������2@ ��87kd?'-§jþ ÿ*&K�w���@��/�3��*><�!Q'"���-á�aLº�Ã�!�����!����>��!����j<M*�/�/1*�$��� �2K��3�d�!Q'"���:�8Oá�aLº�Ã

Y�X"Á:Y-ZV�_@X"V}8a(^�Áº@Ã�¤

ÿ:X?V�[Ã@a(ZÂhº�^�Ä

ñ9T Q 9@S ORQ�Z�=éVWQ=@Z1�/Q�N�\T \/â ê S^L#N5V1� ê O!N��T�O ��<B!�����@�>/13��N*><|§Iþqÿ*&����iV�_@X"V¢!����: 0!��������i���N�i�@�>/�3��;<M*��!����I���>!�� ��< � � þl�iý ��< ��ý m } ©Ký ©cÿl��ý _gÿ9mlý � o Þ �C<���*@A �*�$%*@ ~¦J3���!-_wÿKmDý � o Þ �I$����54��i���K*�<�!Q'"���~/�����!K*�<�/��� "E2�3�����!��C4�����2@ ����<�!����D�@��/�3�� *�<§jþ ÿ*&¢���DÃ@a(Z@!����� �!����c�@��/�3�� *�<�!����D����!����< � � þl�iý ��< ��ý m } ©Ký ©cÿl��ý §jþ ÿ*&pý � oDþ�© �C<M��*@A´ �*�$*� I¦J3���! � oDþ�©��c���D*><w!Q'?���I �*@ "EQ��AB��!Q'�/��1��!D*�<w/��� �2�3�����!��C4���12� ��8Ofg�� ���/�/�'"b"�§<c!����p�@��/�3���*><�§Iþqÿ*&)���hº�Ã�¤j!����: �!����~�@��/�3��*><_wÿKmDý � o Þ ��� !�������AR��!Q'�/�����! �� �7�!������@�>/�3��=*>< � oDþ�©���c!������:AB��!Q'~/�����!�HGeFHGeF ì:9@�/QB�-S�VW�/ToN�T�â�S^Z�â�Z1\Ü ��N�T@��=DN5ORQ�N�V

X'Z1\�]�S^Z1Q�N�S/O> S�?�TDN�Q�NJ\T�\/â'Q9@S�O!S^QB\�â V1�/Q�S^�\�Z1N�S/O�Q9@�/QBT@S^ç�SUZVW�/T@��S�� ê \T@SBN�T��tT@\�Ü�N�T@� ê V 9A=@T/B Ü�\�è�S ê ©c*@ "E���57.������ �20!��������:!P*�<�o ��*@������!�������AB*�7"��/�/��� �2� �*�AB�� ���/4��"3� 6 �8b $c�N4:�� P7"�J�� ��;!����R��3�d�����! > ¸H*�<I!����B��*�4���d"E3�/C�>��'Bb 4J*� �������!��� �2K�� ~!���������!=*><n4:��!��:2@*@�������D!�����!D ��:�@���4��> ;d�����/�*� ��I�� ��pAB*�7"��/QO

> ¸-A « �C b#"�DE& b(� �������>� � ¾)&��J®�F= ê S�I ÿI/�/�!����94:��!��:2@*@�������=!�����!�������AB��A~d������D*�<e!�����>d�*�����7.�J�� ��87 ����!��01�3HG > 3 A~3���!h���(���;!������@��/�3��V�_�X"V~<M*��)§Iþ ÿ'&pO���*@3� ��9�� �7; �*@AR�� ���/���¥��87~��7�¦J�84J!����@���!�����!K4��� id��h!����~���8��7�*><&�R �*@AR�� ���/w4��"3� 6 �h�>���~4J*@ "E4J�:�� ��87�d?'R!���������3�/��@O�F= ê StF ÿI/�/e!����h4���!���2@*��������K!�����!I������AB�:A~d����c*�< !��������! > � �Q7"�J�� ��87���d�*��@�(�eA~3���!�d��&�>����*.4-�5�>!��87k!�*9�c/��J,��C4��>/�� "!���'p$��1!��~!����j�@��/�3���a(Â�`eZ £1U5Y-Z�<�*@��_gÿ9mDý � o Þ O"o /�3��L�>/ �*�3� ��;�� �7�����*� �*@3� ��;!����>!�4:�� �d���3����(70��/�*@ ��H�1 �� �*�AB�� ��>/@4��"3� 6 �����I4J*@ �4-���� ��(7~d"'�!����1� ��3�/��@O���*@!���!����>!Þ 3�/��I¸h�>����/�������!�*~��/�/�!�����4��>!���2�*@�������c<�*@�c$����C4�� Þ 3�/��pµ�>����/������=!�*"*B��� > � ���=�� �4J/�3�7.�87R�1 I1�3~O

¡10

Page 59: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

�J= ê Sm� ÿj/1/p!����P4���!���2@*���������!����>!G�>���iAB�:A~d������;*�<�K1 > G >@> �0L > ¸~�> �7~!�����!I�>���K �*�!&4J*� ����C7"�:���87;!��L��7"�FE!���*� � ���/�'B�>�9�@7:¦J�84J!��1�����I���(�@�K!����p�@��/�3��pÃ@a(Ze<�*@�)§Iþqÿ*&�/T�è ���(�@��<M*@� � oDþ�©����@��/�3���*><w!Q'?���pY�U§W@V�!����>!D���=��3�d"E��3�AB�87�d"' ��< � � þD� ý �M< ��ý�m } ©Ký ©cÿD��ý §jþ ÿ*&pý �*@3� wO&j��!��:��AB�� ��:���I������4J*@ �4-���� ��(7�$���!��;!���������3�/���J= ê SON �������K��3�/��p���� �7./1�:��$���!�����*�������d�/��B4J*�A~d��� ��(E!���*� B*�<D7"��!�����AR�1 ������K��*@�I7"��!�����AR�� ������I�> �7;��3��� "!��F���:���L��� �7B2����@����*� ��K��*�������d���/���!Q'h!�*B4J*@A~d��1 ��I!�����A !�*@2���!����:�8O�Q!;��!��1��3�/C��!��:��!����>!;�§<~��4:��!��:2@*@��' !����8�>!��87��� Þ 3�/��i¶4��> ~�����84-���87��> �*@!������=4��>!���2�*@��'~!����(��!��(7~�� Þ 3�/��I¶�Pj�M$c�6 �*�$i!�����!w!�����*�3�2@�p!����=����/C��!���*� B_@^�Ã@a(^w7"�J�� ��87h�>d�*����8��b!����� ���!I���� ��84-�����J����'�!�*~����*��"�C7"������!������9�R4J*@AR��/��J,N7"�JE!��:��AB�� ��:�D��!���3�4J!�3�����b�*��=!�*h�@7�7~!�*�!����+_wÿ9mlý � o Þ �@��/�3��*�<D��/�/e!����h4���!���2@*��������j!����(��!��(7;�� Þ 3�/1�~µK!����K$���*@/��K/��1��!*�<D7"��!�����AR�1 ������(O�J= ê SRQ ÿI "'�4��>!���2�*@��'�*>< > ¸;!����>!����>�9 �*@!Kd��B4-*@ "E���C7"�:���87¢d"' Þ 3�/��B¶������B!L� 6 �� ��>����7�¦J�84J!����@�N�> �7i���(�@�!����9�@�>/�3��~º@Ã�¤l<M*���§jþ ÿ*&pO�HGe� K2L#Q�SUT�ORN�\�T�O�Q!�����$c��/�/ 6 �*�$� �!�����!p!��������N�>���N7.�§�e������ "! 6 �� �7"�h�> �77"�F�e������ "!~��*@3��J4J���B*><k�>A~d���2@3���!Q'"O;�i�B��*@�� "!h���:����!Q$c**�<p!�����A �> �7���*�$T!����:' 4��� �d���!����8��!��87�$���!����� �*@3��<M�J��AB�:$=*�� 6 Oÿ�/��� �2@3��1��!��C4l�J,�������������*@ ;4��� ~d��9�>����*.4-�5�>!��87p�� ~!����Ime�J,"E�C4J*@ �!�*;AR*@����!����> �*� ���Âh[;ýK��!K���8be�@O 2�O1bn!����B4��>���p<M*��me�(�(<M�I¸~�> �7;¶�*><q!����j������!I�J,���������AR�� "!��1 � �(4J!���*@ �¶�O1µ�O�����~��A~d���2�3���!Q'~�1�j!����:���~���:��*@/����87�!����> 6 �K!�*Nmn�8�(�@�;µ@O� 3�����*����~!�����!8b ���K�� if����� �4��wbg!��������~���k�R��!����� �2H*�<=�J,"E�����������1*� ��B�� �������/C��!��(7P����!�!����� ¢Eh���;£�aK¤�X�W@a�Ek$��������d�*@!����J,�������������*@ �����������A~d���2@3�*@3��~��£1aHd����� �2P�> �����!��FE4J/��i�� �7���4J/���!��C4�bl¤�X�W@aP� �*@3�  *���� �@����d���O �Q �!����1�����!�3���!��1*� wb !�������A~d���2@3��1!Q'i���;AN�>�� ?!J���� ��87nbD!����i��'"��!���A�����84J'��� �2Nd�*@!��iÂ-�M»�V��� �7�Â��ulwVR<�*@��!����~£1ac¤�X�W@aR�����JE�����84J!����@��/�'"b?�K �*@AB�� ��>/?�> �7~�K�@�:��d��>/�4��"3� 6 ��Og�������=�>A~Ed���2@3��1!Q'�$���/�/=d��;������*@/��@�87¢�� ��i4-*@ "!��J,�!pE��@O 2�O¢!�*�!��������2��?!c*�<D����������*@���1!���*@ Nmn�8�(<gED�1 R$����C4��B!����9�-,.������������*@ 4��> �<M*�/�/1*�$��F<=��!K�1�j�����84J�F���87¢���~Â��M»�V�d�3�!K �*�!9�F<=��!K�������Â��ulwV�OÿI�~�� i��AR��*���!L�> "!I���C7"�-E��J�e�84J!~*><I!�����������!h�-,.�������1AR�� "!� � �84J!���*� �¶�O�µ8��b���!=���c����A;��� 6 �87��1 §I���~��!��������8��!�����!����.E��/�'"�� �2~!����p����*.4-�������1 �2�4����>�� i�M������fg��2@3����Bµ(��!�*R�����:�?�FE*@3���/�'N�> �7R�1 �7"������ �7"�� "!�/1'�7"���L��A~d��12�3���!��87p�J,������������1*� ����AB����*��@���c�@�:��'B/���!�!�/��9!����K�� ���/e���:��3�/�!��(O��i�I!����� 6 !�����!*@d��������@��!���*� ��&�>��!������c*@ ��K�� �7"�C4��>!��j!����>!�!����K�� �4J����AB�� .E!L�>/�!L��4J!��C4l*�<ed�*�!�!�*@A~EQ3������������� �2��� �7�!����>! !��������8��3������JEAB�: ?!l*�<q��7"���L��A~d��1!�3���!���*@ �/C�('@�:�=d��J<M*����K���������� �2p���= �*�!!����9*� �/�'B��*@������d�/��K$&�('"O�Q ;!������J,���������AB�� "!������������� "!��87H�� N!�����������������$c��$c*@� 6*@ ~AR*�7"��/���!����� �2��=d�3���/57p$���!���[Lº�Z�� Y-b� �*�!D$���!�� � A+��'"A~Ed�*@/����M�C7"�� "!��§�������J��O ��$c*�d������54 !Q'?�����i*><��C7"�: ?!��F������������h���84J*�2@ ���¥��;ý !�����*� ��p����/C��!��87;!�*B "3�4J/��8�>�9�����L�>�����K*@�4��"3� 6 �8bc$����C4��������H������/�/1�(7TS�V�bUS d����� �2P���@�>���C��d�/1�*@ �»[p l#pwû+VXW^rJrJr§�> �7R!����K*@ ����I����/C��!��(7;!�*~ �*�!� "3�4J/��8���Y-î�ÐBá�ÇJÈQÏCä@â-ä@Î�Ô�Î�Ö"Ê�ÇJËcè�Ó�Ð@É�Ï\Ó§Ç�ÐpÇJÒ�Ñ�Î�ÏMÎxÈMË�Ó�Ð@ÎxÈ�Ô ÉJÐ@ÑJZ:ä@ÉJÐ�ÏCÓ 8

4@ÎQÈQÔ�Ý

�����L�����@b������:/1/��87�$��1!��;!�����d������!S �> �7R��!�����*@������d�/����� "E��!L�� "!��5�>!���*@ ��8Oe�Q p2@�� ����L�>/�b��[S�� Vp�� �!�����AR*�7"��/"��!����� �2K*�<��*@AB�/S]\:V��>���D �*@!��>A~d���2@3�*�3���/�'c����/C�>!��87eO@�l3�!D��!�!J�@4����JEAR�� "!9*><^S^_�V�!�*�!����~���12��"!�*�<9�R���>!�!��:�� @`�¨ , d�d�da` ² ,4��> ;d�����A~d��12�3�*@3��8Oo ��*@������!������=�������j���������� "!��87R������/1'��J,��@4J!�/1'p!������L��AR�=*� AR*.7.��/�����!����� �2@�~$���!��P*@�p$���!���*�3�! � Â�� Y-O � *�!����;�����JE�"��*@3���/�'B4������L��4J!��:�������87i�>A~d���2@3���!Q'h4:�� �d��h�-,.���������87�d"'7"����¦J*@ �4-!����@�;������*�$��� �2��� �������*�$��� �2�<�*@��A~3�/5�H������� � �84�E!��1*� N¸�O ¶���O

b °®³B¸I��µ�v¯õö¹³ ¸�Q �4J3������� "! $c*@� 6 *@ K��'" "!L�ç,~�M����3�������!��54-�w<�*@�e��*�d�3���!w���>���QE�� �2��M��������ÿ9�k© Þ ��µ(��b~���2(?��c@����*@��3� ��§�e4���!��1*� "EQd������872��L��ARAB��!��54:��/�<M*���AN�>/��1��A~�L�:b���!I������3���!��~7"�edN4J3�/�!K!�*��@4�E4J�:���K��3�����/��1 �2@3�����!��C4��� "<M*���AN�>!���*@ ~���� �4J��!������L�>AB�K��'" "E!J�(,P���h3����87�d�*@!��¢<M*@�p!����;/��� �2@3��1��!��C4B7.���L4J������!���*@ ��> �7!����;��3�/����p<M*��h!����;�������������8O��i�Rd���/��������B!�����!~!����N�-,?E������������*@ i*�<I/��� �2�3�����!��C4K�� "<�*@��A;��!��1*� ;d?'�AB�(�� ��K*�<9mn�� "E2�3�����!��C4Ko ��*@������!��1�:�9���K�p��*�������d�/�����!����H�1 H!����h7.�1���84J!��1*� *><�!����I4-�� "!��L�>/��1¥(��!���*� h*><�/��1 �2@3�����!��C4 6 �*�$�/��87"2��D$���!��p!����<M*�/�/1*�$��� �2Kd��� ��J��!��8ý

ã � '? "!J�@4J!��54-�5�> ��~$c*@3�/C7i�����: �7�/������~!���AB�B����$�����!��1 �2��3�/�����4:������'"�1 �29!����D�J��AB�l�1 .<M*@��AN��!���*@ c<M*���7"�F�e������ "!<�*@��A;��/�����A~� *@�c<�*@��7"�F�������� "!I���>���������(O

ã �����G4J*� ���!���3�4-!���*@ %*�<���2��L�>ABAB��!��C4���/����J<M������ �4J�@b�-,.�����������87��� ��H<M*@��AN�>/1��¥��(7P�� �7P �*@ "E���A~d���2�3�*@3��$��('"O

�����R �*@!���*@ P*><9��2��L�>ABAN�>�����~����*@3��J4J�;*�<9/��� �2�3�����!��C46 �*�$�/��87"2��B���p!��"3��p�����"������!��(7P�� <��(�@*���*�<��� �*�!���*@ ¢*�</��� �2@3��1��!��C4 6 �*�$�/��87.2@��d��>����fp<���*@A $����C4�����'" "!L��4J!��C4��� "E<M*���A;��!���*� 94J*�3�/C7pd����J,�!��L�@4-!��87�<M*@� *@ ���*@�D�� �*�!����:�D�����JE4J�F��4k2@�L�>ABA;���D*��k�>����/��C4���!���*@ wOc�����p!Q$=*R�J,����:����AB�� "!��!�����!�$c�N7.���L4J����d��87���d�*��@�~���:��A !�*�d�������!��:���� i!������7"�����84J!���*� wO

gÔº�zYº$²´º:¸j��º:õ� O:ÿihj�!QE��;* 6 �"!J���8OeS�� º@V�º>£ Y�a `e^/ka:Y V�ZMº l@UCÀ(X"a�a8VpX"V�a�Y�a(X�£ka8Z�ºJ`�a(O=o �& !����:�����8b�a� ����@�:������!nm�[�l/C��������o ���J4���/Qbgµ8�@��Ú�O� O:ÿihj�!QE��;* 6 �"!J���8b�(>Exo�©c���� �*.7nbç�> �7-©KO Þ *@3.,nOnÿPA~3�/�!��FE�� ���3�!&7"3���/FE��: ?!���';7"������ �7"�� �4J'����>�����:�8OD�Q �]�^�_8[JaJaJÃ@UCV�W�Y_�ÿ)ú ö ]-à�¢�opo^¡�b#�D�:� ¦-�1 �2�b"¸@�@��µ@O© Ok��q���(O mw� �����L�>���t������d���/��t �*�'"��3��� <M�L�> sr4������(Ot�aJ[9á�a(^�[9á�a:YKY-X"^j£1a¨ÿ�^�º@V/ß[(º@U�Yvu.º�^J£pka(beµ8¡�beµ8���@��O© �+q������ �7�o O2�D/C�@4�����O�o ��*@�����m�:!nm�����:!��> ���/�'"����7e� 3� /C�� �2?��2��@Oe�Q �ûj[8Z�a�Y�Ã�atà"ûISg»xwpw�b/(@3�/1' ©�����2�q�����@bwµ(�@�@��O© �+q������ �7o©KO2§I��2�q��2@��O�o ��*@������!��1�:����  ¡��wO;� �84��� ��§E4��>/e������*@��!8b © ��*�3�����7"� Þ �84������L4����p7��� ���/������Q �7"3���!��������7"�l/5�jmw�> �2@3���� ©-Þ ��mw��b�a Þ mDý�/�2�����/QO 3� ����@EQd��e4J/�����AR*@ "!8O <C�8b¸��@��µ�Oy Þ�ß ÎkÓ§Ñ�Î�ÉpÇJÒ=ÉpÍ�Ó�Ð@âJä�Ó§Ô�ÏCÓ§Ê�é8Ð@ÇJã�ͧÎ�Ñ�â�Î�è@É�Ô�ÎKã�É:Ô9Ç-ÈMÓ§âJÓ�Ð@É-Í�ͧõËIÎxÐçÏ\Ó§ÇJÐ@Î�Ñgè�õ);cÝ�;cÝ(Ø åÎ�Ô�Ó�Ð9ÉlÌ�È�ÇJô�Î�Ê�Ï�Ì�ÈQÇJÌ?Ç:ÔMÉJÍ"× ñ ÉJͧÎLÜ�ÔCä�è�Ë�Ó§Ïq8Ï�Î�ÑcÓ�Ð�Û æ�æ Û(Ý

¡@¡

Page 60: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

© �+q���8b2©KO@§I��2�q��2@��b=�� �7imDO�©c*@����3��8OB&j���p����*������ m��!nm���/��� �2@3�����!��C��3����4q��/Q���� ��>/�'?���N7e� 3� ��N/C�> �2@3���Oh�Q �]�^�_8[JaJaJÃ1�UCV@W>Y9_�ÿKZeá�akl þ S�à"ûIS ø _@V!ÿ�a(^�a(V�[Ja(b?��*��@��A~d�����_D�: ��C4J�@bµ8���@��Oo Oi�l/C�@4����@O ©c*@ ���!��J���� "!��8b�/��� �2@3��1��!��C4)!�����*@������� �> �7 ��>!�3��J��/�/C�> �2@3��>2@� ����*.4-�������1 �2�O �Q &j��AB��!������ ��O©c�����1��!�*�7"*�3�/C� 6 ���8bg�87"��!�*��8bD»�º�ZMX"^�º�£�Sgº@V@W�X"º8W�aB]�^�_8[Ja:YLY9�UCV@W �B»ISg] ¢�opopo�bp����2�����¸@¸�µ)z�¸@¶@¸�O � ������ �2����QE5_D����/C��2�b¸@���@��Oog*@/�/C���J7�©KO��> �7i��O � �>2�O!{�aJº@ÃW�KV�^�UJ}(a(VG]�á�^�º>Y�a � Z�^�X"[!�Z�X?^�a]|�^�º@ÂpÂhº�^�O�© � me��me�(4J!�3�������*@!����8O#©c�� "!��:�I<�*@��!����� !�3�7"'B*><qmg�� �2�3���2��k�> �7~�Q .<M*@��AN��!���*@ wb�µ(�@�W0�O©KO/§I�>2�q��2��@ODûKV�º>£ Y�aKY V�Z�ºl�UCÀ8X"apº@X"Z�_@Âpº@Z�UCÀ8X"apÃ@XXu�_@^9�Z�XçW�º@U5Y-ODoq�&0!����������(ba� ��1����������!nm� �l/C��������o ���J4���/Qb�¸@���@��O©KOB§I�>2�q��2��½�> �7 © OB�+q���8O &K�0*�d����:���@��r4~}��*t7"������*>E�����1�(7��@7"�:�t/��1 �2@3Amj���!��C4�����q� ��3�� <M*@��AN��/���¥8��r4 }��* "3�AN�2@�J��A�m��!��C4���7"*~����*�4J�:���L�>AB�� "!�*B7��p/�mj1 �2@3���Ow�Q �ûj[8Z�º�Y�Ã@_ú�ú�ú þ V�[J_@V�Z�^�_�u�º@^�º�_;]�^�_([Ja�YJY�º@Âpa(V�ZM_ ø _�ÂUueX"ZMº�[8UC_�V�º�£Ã@ºKS!k� V@W@X"ºK]�_@^:ZMX�W�X?a:Y�º|��]-tX�D]��-t � w��R�@b@og*@��!�*KÿI/���2@���@bµ8���uÚ�O©KO�og*�/1/C�>�L7B�> �7 � ��2~��O�ÿhOMúLV!ÿ�_�^�Âpº@Z�UC_@V^� ü º�Y�aJà �� V�ZMº lº@V�à � a8Âpº@V�Z�UC[:Y9p l�_�£�X"Âha�ú�� ÷ X"V�Ã@º�Âha(V�Z�º�£FY-O2© � mn�&mn�84�E!�3�������*�!����K��Oeµ8¶�O/©c�� "!�����<M*@�c!���� � !�3�7"'R*�<Dmw�� �2@3��>2@��� �7B�Q .<M*@��AN��!���*@ wb�µ(�uÚ�c�O��O � �>2��� �7H��Oe�P����*�$�O � V�Z�º@[(Z�UC[�à#á�aJ_@^ �pû ÿ�_�^�Âhº>£úLV�ZM^�_8Ã�X?[(Z�UC_@V�O�©c�: ?!����c<M*��D!���� � !�3�7"'~*><�mw�� �2@3��>2@�9�> �7�Q "<M*���A;��!���*� wb � !J�� "<M*��L7 a� ��1����������!Q'?b�µ8���@��OmDO�� ���� ���q�����@O þ £pka(Âha(V�Z\Y Ã@a´Y V�Z�ºl�a´Y-Z�^�X"[(ZMX"^�º>£1a(O�j/��1 �4 6 ���1� 6 beµ(�p����Oo O��q�>���� ����� ��� ~�� �7p��OW(�h�����"�1 ��� wO ÿ� �*� "EQ��*8¦J�(4J!����@�K7"�JE���� �7"�� �4J'#���>�������8O¢�Q +]�^�_8[JaJaJÃ@UCV�W�Yi_�ÿiZfá�a;�@Zfá ø _�V^�ÿ�a(^�a(V�[JaB_�V�û�upu�£1UCaJÃB»�º@Z�X"^�º�£wSgº@V@W�X"º8W�a�Y�b �P�>������ �2�!�*@ &pO ©KO1b�µ8���pc�O��Oc� *@�������N�� �7 o O��;��/�/����8O ÷ _@^�Âpº�£�U�Y�Âha:Y�Y V�Z�ºl�UCÀ(X?a:Yu�_@X"^~£�aBZ�^�º@U\ZMa(Âha(V�Zjº@X"ZM_�Âhº�ZMU\À8X"a�Ã�X�£1º@V�W@º(W@a;V�º�ZMX"^�a:£1O§��:��AOq�:�8bwµ(�@����O

¡p�

Page 61: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

�i�����K���a����5���/���E���5�

��� �¢¡¢�¢£ �a�i¤�¡��E¡¢¥ ¦£ ¡a� �¢¡a§)¨© ��ª�« £ �a¨K¬�¡a��¥�¨K­a� © �5��5���)¨s¤ �K­¢­a¨E��¡a§)�K��§)¨®-¡a¨E¯°����£ ¬�¬�¡�±

² ³ ± ´ ´�µ���¶��¸· ���a¨K­¢¹º

�5­a�)����¨K¥�¤�����E»E� © �

¼5¥ ¡¢½��K��� ¡¢�a¾s¤�¡�µ�¡a��¨®-¡a¨s¶�¨K¥ �a¨E¯°¨K��¡a¨s¿ ³² ³ ± ´ ´�µ���¶��¸· ���a¨K­¢¹º

¶�¡¢ªÀ��¥��K�¢�a¨¯°���5��»E¯!�+Á��5�

��� �¢¡¢�¢£ �a�i¤�¡��E¡¢¥ ¦£ ¡a� �¢¡a§)¨ © ��ªÀ« £ �a¨K¬�¡a��¥�¨K­a� © �5��5���)¨s¤ �K­¢­a¨E��¡a§)�K��§)¨®-¡a¨E¯°����£ ¬�¬�¡�±

² ³ ± ´ ´�µ���¶��¸· ���a¨K­¢¹º

®-¡¢�a�µ�������»E�E�E�

��� �¢¡¢�¢£ �a�i¤�¡��E¡¢¥ ¦£ ¡a� �¢¡a§)¨© ��ª�« £ �a¨K¬�¡a��¥�¨K­a� © �5��5���)¨s¤ �K­¢­a¨E��¡a§)�K��§)¨®-¡a¨E¯°����£ ¬�¬�¡�±

² ³ ± ´ ´�µ���¶��¸· ���a¨K­¢¹º

à �������K���a� Ä ��¨K���a��­¢¡¢¥ ¡aÅ ¨K­a�)����¨K¥�¤���� Ä ­a�K¥�§K¡aÅ � ¡¢ªÀ��¥��K�¢�a¨)Ä ªÀ��¥ �a�KªÀ¨K¦¥ ¡aÅ�½¡¢�a� Ä « ¡¢�����K­¢­¢¡aÆ%Ç[¡¢­a§)Ä §K¥ ��Ä ¡¢�

ÈvÉ ÊÌË Í�Î�Ï�Ë��¥Ð�¢Ñ��5«�¨K«��K�sÒ+�5���K«������-¨^Ó�£�¨K­¢¡¢�a¨K�¢¡¢½��^�K½�¨K­¢£�¨)Â�¢¡a��¥2� ÔU�¢Ñ��!«��K��Ô�����ªÀ¨K¥�§)�°� ÔX¨°¤ �K«��K¥�¤ �K¥�§K¹¨K¥�¨K­¢¹����K�[� Ô'���a¨K­¢¡a¨K¥!�¢Ñ�¨K�A��£ ¥��s¡¢¥!�����¢Ñ°¨i¥���¥�­a�KÕ ¡a§)¨K­¢¡a���)¤J¨K¥�¤C¨�­a�KÕ ¡a§)¨K­¢¡a���)¤�ªÀ� ¤ �)ÄM�i�)� £ ­¢�a�� Ñ��)¤!­¢¡¢¦Ñ �À��¥°�¢Ñ��F§)��¥ �¢��¡¢� £ �¢¡a��¥@� Ô-�¢¹«��)�Ð� Ô­a�KÕ ¡a§)¨K­�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥i�a��«�¨K��� ¡¢¥ ¦�ÄÖÌ× Ë Í�Ø�Ù Ú Ï�Ë Û Ø ×���À¡a�ÐÒ5¡a¤ �K­¢¹Ü¨)��� £ ªÀ�)¤I�¢Ñ�¨K����¡a§KÑ(§)��ª�« £ �a¨K�¢¡a��¥�¨K­­a�KÕ ¡a§)��¥��5Ô�����ªÝ¨�Ô £ ¥�¤ ¨KªÀ�K¥ �a¨K­7§)��ª�«���¥��K¥ �7� ÔM���K­¢¡a¨K� ­a�A«�¨K��� ¡¢¥ ¦!¨K��§KÑ ¡¢�a�)§K�¢£ ���)�/¨K¥�¤-�¢Ñ�¨K�n­a�KÕ ¡a§)¨K­n¡¢¥�Ô�����ªÀ¨K�¢¡a��¥Þ§)¨K¥Þ��¥ ­¢¹ßÑ�¨K½��à���K¥��)Ô ¡a§K¡a¨K­¸�)Ô�Ô��)§K�a�@��¥«�¨K��� ¡¢¥ ¦�ÄE¶�¡¢¥�§)�/�¢Ñ��/���K¦¡¢¥ ¥ ¡¢¥ ¦F� ÔnÒ+����áF��¥À� ��� ¨)¤ §)��½��K��¨K¦��Ы�¨K��� ¡¢¥ ¦â· ã�K¥����K¥T± ä å å ¨)Åi± ä å å�� º�Å��¢Ñ��á��K¹U¡a��� £���Ñ�¨)�n���)�K¥UÑ���Òæ�a�sªÀ¨Ká��-�)Ô�Ô��)§K�¢¡¢½��'£����-� Ô­a�KÕ ¡a§)¨K­s¡¢¥�Ô�����ªÀ¨K�¢¡a��¥�Ä���¥J�¢Ñ ¡a�5«�¨K«��K�iÒ+�ç« £ �s�¢Ñ��)���¨)��� £ ª�« �¢¡a��¥��M�a�A�¢Ñ��E�a�)� ��� ¹[¨)¤ ¤����)��� ¡¢¥ ¦v�¢Ñ��sÔ���­¢­a��Ò+¡¢¥ ¦�Ó�£��)� �¢¡a��¥���è)�a��Ò5Ñ�¨K���KÕ �a�K¥ ��� Ñ���£ ­a¤i¨E­a�KÕ ¡a§)��¥i����¢��£�� �a�)¤-Ô����M«�¨K��� ¡¢¥ ¦�é�ê�Ñ�¨K�~¡a���¢Ñ���¥��)¨K�n§)��¥ �¢��¡¢� £ �¢¡a��¥� Ô ­a�KÕ ¡a§)¨K­�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥i�a�i��½��K��¨K­¢­�«�¨K�����s� £�§)§)�)����éêR�5« ���)���K¥ �'Ñ��K���/�¢Ñ��/���)� £ ­¢�a�v� Ô7¨/« ���K­¢¡¢ª�¡¢¥�¨K��¹

�K½�¨K­¢£�¨K�¢¡a��¥°� Ô��¢Ñ��v¡¢¥ �a�K��« ­a¨K¹X���K�¢Ò+�)�K¥ë­a�KÕ ¡a§)¨K­i¨K¥�¤¦��¨Kª�ªÀ¨K�¢¡a§)¨K­M¡¢¥�Ô�����ªÀ¨K�¢¡a��¥U¡¢¥U«�¨K��� ¡¢¥ ¦ç���a¨K­¢¡a¨K¥U£�� ¡¢¥ ¦¨'����� £�� �M«�¨K��� ¡¢¥ ¦Ð� ¹�� �a�Kªà��¨)���)¤ç��¥Ð¨K¥ç¡¢¥�§K���KªÀ�K¥ �a¨K­¨K« « ��� ¨)§KÑ+�a�+� Ñ�¨K­¢­a��Òì� ¹¥ �a¨)§K�¢¡a§v¨K¥�¨K­¢¹�� ¡a��Ä���Ñ��v� ¹���Â�a�KªÜ§)¨K¥U��£ ¥U¡¢¥U�����¢ÑX¨�¥���¥� ­a�KÕ ¡a§)¨K­¢¡a���)¤ç¨K¥�¤ç¨�­a�KÕ ¡a§)¨K­¢¡a���)¤°ªÀ� ¤ �)Ä © ¨K���)Ô £ ­]¨K¥�¨K­¢¹�� ¡a�0� Ô��¢Ñ��J���)� £ ­¢�a�� Ñ���Ò+���¢Ñ�¨K�'§)��¥ �¢��¡¢� £ �¢¡a��¥Ð� Ô�­a�KÕ ¡a§)¨K­p¡¢¥�Ô�����ªÀ¨K�¢¡a��¥��a�«�¨K�����[� £�§)§)�)���7¡a�AªÀ���������K­a�)§K�¢¡¢½��-�¢Ñ�¨K¥0§)��ª�ªÀ��¥ ­¢¹¨)��� £ ªÀ�)¤ Åv�¢Ñ £�����¨K¡a� ¡¢¥ ¦ç�¢Ñ��'«�¨K��¨K­¢­a�K­M¡a��� £��)�A� Ô�Ñ���Ò�a�5« ����ªÀ���a�v¨�ªÀ�����v�)Ô�Ô��)§K�¢¡¢½���¡¢¥ �a�K¦��¨K�¢¡a��¥^���K�¢Ò+�)�K¥

«�¨K�����K����¨K¥�¤ç­a�KÕ ¡a§)��¥��i¨K¥�¤çÑ���Òí�a�Ф �K½��K­a��«Ð���K�¢�a�K�­a�KÕ ¡a§)��¥��'Ô�����«�¨K��� ¡¢¥ ¦�Äî ï�ð × Ë Î�Ï�Ë Û Ï�ñ Î�Í�ÊÌÛ × ò7ó ô�õ Û Ï�Ø × Ê¶�¹¥ �a¨)§K�¢¡a§O­a�KÕ ¡a§)¨K­]¡¢¥�Ô�����ªÀ¨K�¢¡a��¥(¦��K¥��K��¨K­¢­¢¹#Ô��)�)¤ �«�¨K��� ¡¢¥ ¦(� ¹�� �a�KªÀ�0¤�¡a� �¢¡¢­¢­a�)¤I¡¢¥àö%÷ ø ù)ú û¢üKý�þ�ÿ�� �%ú û��¢þ��� ÿ%ú��+üKöÄ�¶�£ ��§)¨K�a�K¦�����¡¢¬�¨K�¢¡a��¥�¡a�-¨UÔ�����ªÀ¨K­/� «��)§K¡aÔ ¡a§)¨)Â�¢¡a��¥[� Ô�¨n« ���)¤�¡a§)¨K�a�n« Ñ ��¨)��¨K­E§)��¥ �a�KÕ �~¡¢¥-�a�K��ªÀ��� Ô%�¢Ñ���¢¹«���� Ô-¨K��¦£ ªÀ�K¥ �a�U� ¹¥ �a¨)§K�¢¡a§)¨K­¢­¢¹I���K­a�)§K�a�)¤+� ¹X�¢Ñ��« ���)¤�¡a§)¨K�a�J�K¥ �¢��¹à·��)Ä ¦�Ä��¢Ñ���½��K�����¢ûi���K­a�)§K�a�ÀÔ����ë¨� £ �����)§K�5�5µí¨K¥�¤F¨K¥O�������)§K�/�5µ�º�Ä��s�KÕ ¡a§)¨K­^Ô ��¨KªÀ�)�§)��ª�ªÀ��¥ ­¢¹[¡¢¥�§K­¢£�¤ �)è�¡aÄ ºA¥ £ ª����K��� ÔA���K­a�)§K�a�)¤[¨K��¦£�ªÀ�K¥ �a��Å�¡¢¡aÄ º^� ¹¥ �a¨)§K�¢¡a§+§)¨K�a�K¦�����¡a�)�i� ÔE�¢Ñ��K¡¢�s«�� ��� ¡¢� ­a����)¨K­¢¡¢¬�¨K�¢¡a��¥Ð· �5µ'Å�µ�µ'Å��K�a§)Ä º�Å�¡¢¡¢¡aÄ ºE­a�KÕ ¡a§)¨K­�§)��¥�� �¢��¨K¡¢¥ �a���¥+�¢Ñ��i¨K��¦£ ªÀ�K¥ �����)¨K­¢¡¢¬�¨K�¢¡a��¥!·��)Ä ¦�Ä��¢Ñ���« ���K«�� � ¡¢�¢¡a��¥Ñ��)¨)¤�¡¢¥ ¦à¨[µ�µ;§)��ª�« ­a�KªÀ�K¥ �¢º�Å/¨K¥�¤X¡¢½�Ä º^�¢Ñ��C¨K��¦£�ªÀ�K¥ ��Ô £ ¥�§K�¢¡a��¥�¨K­p����­a�)Ä~���¢Ñ��K�n�¢¹«��)�'� Ô�� ¹¥ �a¨)§K�¢¡a§E¡¢¥�ÂÔ�����ªÀ¨K�¢¡a��¥Þ�¢Ñ�¨K�ܨK��� ¨K­a��� Ô���£ ¥�¤ß¡¢¥ � ¹¥ �a¨)§K�¢¡a§­a�KÕ ¡a§)��¥��s¨K���)è�¨K��¦£ ªÀ�K¥ ����« �¢¡a��¥�¨K­¢¡¢�¢¹�Å�½��K���ë§)��¥ �¢����­aŨK£ Õ ¡¢­¢¡a¨K��¹À���K­a�)§K�¢¡a��¥�Å~����¤ �K�7§)��¥�� �¢��¨K¡¢¥ �a��Å~�K�a§)Ä~��¥��¢Ñ�����¢Ñ��K�ÐÑ�¨K¥�¤ Åç§)��­¢­a� §)¨K�¢¡a��¥� ��¨)���)¤]­a�KÕ ¡a§)¨K­^¡¢¥�Ô�����ªÀ¨)Â�¢¡a��¥æ¡a�ë��¥ ­¢¹°��¨K���K­¢¹°« ����½¡a¤ �)¤C� ¹@§)��ª�« £ �a¨K�¢¡a��¥�¨K­­a�KÕ ¡a§)��¥���Å�¨�¦�¨K«!� Ô �a�K¥+­a¨KªÀ�K¥ �a�)¤5¡¢¥+����� £�� ��«�¨K��� ¡¢¥ ¦� ¹�� �a�Kª;¤ �K½��K­a��« ªÀ�K¥ �aÄ� ¥ £ ª����K�]� ÔÀ� ¹¥ �a¨)§K�¢¡a§0§)��ª�« £ �a¨K�¢¡a��¥�¨K­^­a�KÕ ¡aÂ

§)��¥��[¨K���U¥���Ò+¨)¤ ¨K¹���¨K½�¨K¡¢­a¨K� ­a�U�a���¢Ñ��U�5�Eµ@§)��ªÀª�£ ¥ ¡¢�¢¹�Äx��ª�«������a¨K¥ � �KÕ�¨Kª�« ­a�)� ¨K��� � �+� © »· µ���� §K�a�K�/± ä å���º�Å © ��ª��s�KÕз�Á���¡a� Ñ ªÀ¨K¥Àü)û�ú�����± ä ä�� º�ŵ��5�i���E»Ü· ��£ ¡¢ª�¹æü)û/ú����7± ä ä å�º�Ä���Ñ��)���U­a�KÕ ¡a§)��¥��¨K���^��¨)� ¡a§)¨K­¢­¢¹FÑ�¨K¥�¤ Â�§K��¨)Ô �a�)¤À� ¹O�KÕ «��K���5­a�KÕ ¡a§)��¦��¨)« Ñ��K����Åv¨K¥�¤C�¢Ñ��K¡¢��¥�¨K�¢£ ��¨K­�« £ ��«�� ����¡a�U�a�!« ����½¡a¤ �¦��K¥��K��¨K­�« £ ��«�� ���)Å-¤ ��ªÀ¨K¡¢¥� ¡¢¥�¤ �K«��K¥�¤ �K¥ ��� ¹¥ �a¨)§K�¢¡a§¡¢¥�Ô�����ªÀ¨K�¢¡a��¥�Å�§)��½��K��¡¢¥ ¦X�¢Ñ��vªÀ� � �iÔ ���)Ó�£��K¥ �i�K¥ �¢��¡a�)�¨K¥�¤�Ô ��¨KªÀ�)��Ä���¥��¢Ñ��U���¢Ñ��K�AÑ�¨K¥�¤ Å «�¨K��� ¡¢¥ ¦C� ¹�� �a�KªÀ�

²��

Page 62: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

� Ô �a�K¥!§)��ª�« ­a�KªÀ�K¥ ��¦��K¥��K��¨K­�­a�KÕ ¡a§)��¥��EÒ5¡¢�¢Ñë§)����« £���¤���¡¢½��K¥�Å�¨K£ �a��ªÀ¨K�¢¡a§)¨K­¢­¢¹CÑ�¨K��½��)� �a�)¤J� ¹¥ �a¨)§K�¢¡a§�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥�· �'�)¤ �K��¡a§K¡Jü)û�ú�����± ä ä å���Å5����¡a��§)� ����´ ´ ± Å������Ñ���¥��K¥���´ ´�� º�Ä��5£ �a��ªÀ¨K�¢¡a§+¨)§)Ó�£ ¡a� ¡¢�¢¡a��¥�� Ôs� £ ��§)¨K�a�K¦�����¡¢¬�¨K�¢¡a��¥æÔ ��¨KªÀ�)�+¨K­¢­a��Ò+�+� ¹�� �a�KªÀ�5�a�C¨)§)§)�)���Ñ ¡¢¦Ñ ­¢¹À§)��¥ �a�KÕ �'¤ �K«��K¥�¤ �K¥ ��§)��¥�� �¢��£�§K�¢¡a��¥���Å��a��Ô ¡¢­¢­p¡¢¥«�� ��� ¡¢� ­a�A­a�KÕ ¡a§)¨K­n¦�¨K«��/¨K¥�¤+�K½��K¥ �¢£�¨K­¢­¢¹^���K­¢¹ë��¥!Ô ���)ÂÓ�£��K¥�§K¹-¡¢¥�Ô�����ªÀ¨K�¢¡a��¥v�a���¢£ ¥��n�¢Ñ��n���K­a¨K�¢¡¢½���¡¢ª�«�¨)§K�n� Ô� «��)§K¡aÔ ¡a§sÔ ��¨KªÀ�)�'· © ¨K������­¢­Eü)û�ú����~± ä ä å�º�Ä�s�KÕ ¡a§)��¥I§)��½��K��¨K¦��v¡a�7£�� £�¨K­¢­¢¹X���K¦�¨K��¤ �)¤ë¨)�A�¢Ñ��

ªÀ¨K¡¢¥U«�¨K��¨KªÀ�K�a�K�/¨)Ô�Ô��)§K�¢¡¢¥ ¦ç£����-� Ô�­a�KÕ ¡a§)¨K­�¡¢¥�Ô�����ªÀ¨)Â�¢¡a��¥]Ô����7«�¨K��� ¡¢¥ ¦�Ä��+��Ò+�K½��K��Å �¢Ñ��s���)¨K­/§)��ª�«�¨K��¨K�¢¡¢½��¡¢ª�«�¨)§K�ç� Ô��¢Ñ���û ���Aüз ��¨K�¢Ñ��K�v�¢Ñ�¨K¥C�¢Ñ���ªÀ�K���ÀÓ�£�¨K¥�Â�¢¡¢�¢¹º5� Ôp­a�KÕ ¡a§)¨K­�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥^Ñ�¨)�n���)�K¥ë���K­a¤ ��ªÜ¤�¡a��§K£������)¤ Ä���£ �E���)� £ ­¢�a��� Ñ���Ò°�¢Ñ�¨K�p�¢Ñ��/§)��¥ �¢��¡¢� £ �¢¡a��¥À� Ô½�¨K��¡a��£��-­a�KÕ ¡a§)¨K­/¡¢¥�Ô�����ªÀ¨K�¢¡a��¥F�¢¹«��)�-�a��«�¨K�����X� £�§)§)�)����¡a��¥�����£ ¥ ¡aÔ�����ªÀÄ���Ñ��7�KÕ «��K��¡¢ªÀ�K¥ �EÔ�� §K£����)����¥[¨«�¨K���¢¡a§K£ ­a¨K��� £ �����K�/� Ô��¢Ñ��s¡¢¥�Ô�����ªÀ¨K�¢¡a��¥J¨K½�¨K¡¢­a¨K� ­a�7¡¢¥� ¹¥ �a¨)§K�¢¡a§�­a�KÕ ¡a§)��¥��IÂ+�¢Ñ��]���K« ���)���K¥ �a¨K�¢¡a��¥Ý� Ô[µ�µ§)��ª�« ­a�KªÀ�K¥ �a��¡¢¥0­a�KÕ ¡a§)¨K­[Ô ��¨KªÀ�)�ÀÂ7�a�)� �a�)¤���¥��¢Ñ���a¨)� á:� Ô0µ�µ'Â�¨K�¢�a¨)§KÑ ªÀ�K¥ �aÄX��Ñ��@���)¨)����¥ Ô����à�¢Ñ ¡a�§KÑ���¡a§)��¡a�E�¢Ñ�¨K���¢Ñ ¡a�E« ¡a�)§)�i� Ôp¡¢¥�Ô�����ªÀ¨K�¢¡a��¥ë� §)§K£ « ¡a�)�¨�§)�K¥ �¢��¨K­A¨K¥�¤+¤ ��ª�¡¢¥�¨K¥ ��«�� � ¡¢�¢¡a��¥+¡¢¥!�KÕ ¡a� �¢¡¢¥ ¦^­a�KÕ ¡a§)��¥���Ä��'���/¡¢¥�� �a¨K¥�§)�5¡¢¥Ð�¢Ñ��5���a¨K­¢¡a¨K¥Ðµ��5�i���E»I­a�KÕ ¡a§)��¥�Å%ªÀ�����A�¢Ñ�¨K¥C��¥����¢Ñ ¡¢��¤+� Ô�½��K���!Ô ��¨KªÀ�)�s§)��¥ �a¨K¡¢¥«�� � ¡¢�¢¡a��¥��s���)¨K­¢¡¢¬��)¤+� ¹I¨iµ�µ'Å'¨K¥�¤+�¢Ñ ¡a�7«��K��§)�K¥ �a¨K¦����¨K¡a���)�@£ « �a�ß�¢Ñ��ì¥��)¨K�T�a���a¨K­¢¡¢�¢¹ ¥���£ ¥� Ñ��)¨)¤ �)¤Ô ��¨KªÀ�)��Ä� ��Ø�É Ú ÊÌË� �Î�Í�ÊÌÛ × ò Ø"! Ö Ë Î ó Û Î ×��Ñ���¦��K¥��K��¨K­[¨K��§KÑ ¡¢�a�)§K�¢£ ���C� Ô/�¢Ñ������a¨K­¢¡a¨K¥0«�¨K��� ¡¢¥ ¦� ¹�� �a�Kªí£����)¤�Ô����'�a�)� �¢¡¢¥ ¦ë¨)¤�Ñ��K���)�E�a�5�¢Ñ��iÔ���­¢­a��Ò5¡¢¥ ¦« ��¡¢¥�§K¡¢« ­a�)��è-±�ºUªÀ� ¤�£ ­a¨K�X¨K« « ��� ¨)§KÑ��a�Ы�¨K��� ¡¢¥ ¦�Å#� º£ ¥�¤ �K��� «��)§K¡aÔ ¡a�)¤R��£ �¢« £ �!· Ò5Ñ��K¥��K½��K�Ð���)Ó�£ ¡¢���)¤�º�Åç¿�º§)¨K£ �¢¡a��£��M£����/� Ô~­a�KÕ ¡a§)¨K­p¡¢¥�Ô�����ªÀ¨K�¢¡a��¥�Å�¦��K¥��K��¨K­¢­¢¹i���)Â�������a�)¤7�a�A¡¢¥À����¤ �K�n�a�A���)Ô ¡¢¥��/¨K¥�¤�$a���7Ô £ ���¢Ñ��K�7� «��)§K¡aÔ ¹¨K¥�¨K­¢¹����)�ܨK­¢���)¨)¤�¹x« ��� ¤�£�§)�)¤ ��¥ �¢Ñ��;��¨)� ¡a�2� Ô¦��¨Kª�ªÀ¨K�¢¡a§)¨K­ç¡¢¥�Ô�����ªÀ¨K�¢¡a��¥�Ä'��Ñ��)����« ��¡¢¥�§K¡¢« ­a�)�U£ ¥�¤ �K��­¢¡a�[���¢Ñ��K�5�¢¹« ¡a§)¨K­������ £�� ��«�¨K��� ¡¢¥ ¦�¨K��§KÑ ¡¢�a�)§K�¢£ ���)�· © Ñ�¨K¥�� ¤%��´ ´ ± Å�����¡a��§)� �s¨K¥�¤ © ¨K������­¢­���´ ´�� º�Ä��Ñ�� � ¹�� �a�Kª §)��¥�� ¡a� �a� � ÔH¡aÄ º © �5¼5�&�� ���· �'�)¤ �K��¡a§K¡iü)ûAú�����± ä ä å ¨Kº�Å'¨i��¨K�¢�a�K��¹0� Ô5Ô ¡¢¥ ¡¢�a�[� �a¨K�a�¨K£ �a��ªÀ¨K�a¨¸Ô����]¥���¥� ���)§K£ ��� ¡¢½��0�a�KÕ �I���K¦ªÀ�K¥ �a¨K�¢¡a��¥·�ù'��÷���(�� � ý~º�Å ¨K¥�¤�¡¢¡aÄ º��)�5»E�5��· �s�K¥�§K¡Eü)û�ú����*��´ ´ ±�º�Ũ+ ü,�Aü'� + ü'��ù,�.-�ø ú�ö%ü + ¨K¥�¨K­¢¹����K��� Ô��¢Ñ��[Ô £ ­¢­���¨K¥ ¦��[� Ô¡¢¥ �¢��¨)Â����K¥ �a�K¥ �¢¡a¨K­�Ô £ ¥�§K�¢¡a��¥�¨K­M���K­a¨K�¢¡a��¥��A·��)Ä ¦�Ä�� £ �����)§K�aÅ�������)§K�aÅ�ªÀ� ¤�¡aÔ ¡a�K��Åp§)��ª�« ­a�KªÀ�K¥ �aÅp�K�a§)Ä º�Ä © �5¼5�&�� ������)Ó�£ ¡¢���)�ݨܪ�¡¢¥ ¡¢ª�£ ª � Ô ­a�KÕ ¡a§)¨K­¸á¥���Ò5­a�)¤�¦��)è­a�Kª�ªÀ¨)Å�«�¨K����� Ô�� «��)�)§KÑ�¨K¥�¤�ªÀ����« Ñ�� Â�� ¹¥ �a¨)§K�¢¡a§sÔ��)¨)Â�¢£ ���)��Ä��)�5»E�5�J¡¢¥�§K­¢£�¤ �)�E¡¢¥+�¢£ ��¥+�¢Ò+�5ªÀ¨K¡¢¥!§)��ª�«�� Â¥��K¥ �a��è�· ¡aÄ ºT¨ © �����/�+�K«��K¥�¤ �K¥�§K¹xÁ���¨Kª�ªÀ¨K�í� Ô

���a¨K­¢¡a¨K¥0�· ¡¢¡aÄ ºç¨^� ¹¥ �a¨)§K�¢¡a§/­a�KÕ ¡a§)��¥]� Ô#12��³ Å ��´ ´À� £ ��§)¨K�a�K¦�����¡¢¬�¨K�¢¡a��¥OÔ ��¨KªÀ�)�+Ô����i¥���£ ¥���Å�½��K�����+¨K¥�¤J¨)¤ Â���)§K�¢¡¢½��)�ߤ �K��¡¢½��)¤:Ô ����ª �¢Ñ��(���a¨K­¢¡a¨K¥Þµ��5�i���E»� ¹¥ �a¨)§K�¢¡a§à­a�KÕ ¡a§)��¥Þ· ��£ ¡¢ª�¹ ü)ûIú�����± ä ä å�º�Ä���Ñ���)�5»E�5� © �����UÁ���¨Kª�ªÀ¨K�A¡a�-Ô�����ªÀ�)¤i� ¹31�± ´ ´v��£ ­a�)�· ¡¢ª�« ­a�KªÀ�K¥ �a�)¤À¨)�vÔ ¡¢¥ ¡¢�a�^� �a¨K�a�U¨K£ �a��ªÀ¨K�a¨Kº�§)��½��K��¡¢¥ ¦ªÀ¨,�����v� ¹¥ �a¨)§K�¢¡a§A« Ñ��K¥���ªÀ�K¥�¨)Å 4A¨K¥�¤+����¦�¨K¥ ¡¢¬��)¤5¡¢¥ �a�ö%û ÿ%÷ ù)û¢÷�ÿ%ú���� �.-�ø ú�ö%ü + ÿ%÷��¢üKö ¨K¥�¤5�¢ü'6��¢ù)ú���� �.-�ø ú�ö%ü +ÿ%÷��¢üKöÄE�)�5»E�5�;¨)¤ ��« �a��¨!� ­¢¡¢¦Ñ �¢­¢¹ � ¡¢ª�« ­¢¡aÔ ¡a�)¤À½��K��Â� ¡a��¥�� ÔE�¢Ñ��&���5¯!»¸¨K¥ ¥����a¨K�¢¡a��¥]��§KÑ��KªÀ�^· �s�K¥�§K¡/ü)ûú����#��´ ´ ´�º�Å�Ò5Ñ��K���XÔ £ ¥�§K�¢¡a��¥�¨K­/���K­a¨K�¢¡a��¥��+¨K���çÑ��)¨)¤ Â��¨)���)¤�¨K¥�¤-Ñ ¡a�K��¨K��§KÑ ¡a§)¨K­¢­¢¹ë����¦�¨K¥ ¡a���)¤5�a�5ªÀ¨Ká���« ��� ½¡a� ¡a��¥ Ô����2£ ¥�¤ �K��� «��)§K¡aÔ ¡a�)¤ ���K« ���)���K¥ �a¨K�¢¡a��¥��Ý� ÔÑ ¡¢¦Ñ ­¢¹C¨Kª�� ¡¢¦£���£��5Ô £ ¥�§K�¢¡a��¥�¨K­7¨K¥�¨K­¢¹����)��Ä%��Ñ ¡a�/Ô��)¨)Â�¢£ ���v¨K­¢­a��Ò+�n�)�5»E�5�Ð�a�/�a¨)§Ká­a�v§)¨)���)��Ò5Ñ��K���'­a�KÕ ¡a§)¨K­¡¢¥�Ô�����ªÀ¨K�¢¡a��¥F¡a�-¡¢¥�§)��ª�« ­a�K�a�)ÅA����Ò5Ñ��K���ÐÔ £ ¥�§K�¢¡a��¥�¨K­���K­a¨K�¢¡a��¥��^§)¨K¥ ¥����7���À¤�¡a��¨Kª�� ¡¢¦£�¨K�a�)¤!§)��¥�§K­¢£�� ¡¢½��K­¢¹·��)Ä ¦�Ä�¡¢¥^�¢Ñ��v§)¨)���v� Ôp�¢Ñ��v¨K��¦£ ªÀ�K¥ ��½���Äp¨)¤7��£ ¥�§K��¤�¡a��Â�¢¡¢¥�§K�¢¡a��¥ º�Ä��98)§)��¥�Ô ¡a¤ �K¥�§)�à��§)������:�¡a�°¨)����� §K¡a¨K�a�)¤Ò5¡¢�¢ÑÀ����ªÀ�/� Ô �¢Ñ��E¡a¤ �K¥ �¢¡aÔ ¡a�)¤i¤ �K«��K¥�¤ �K¥�§K¹v���K­a¨K�¢¡a��¥���a��¤ �K�a�K��ª�¡¢¥��!¨+« ­a¨K£�� ¡¢� ¡¢­¢¡¢�¢¹F��¨K¥ á¡¢¥ ¦O¨KªÀ��¥ ¦O¤�¡aÔ�ÂÔ��K���K¥ ��«�� ��� ¡¢� ­a�s¨K¥�¨K­¢¹����)��Ä��¥ë�)�5»E�5�sÅ�­a�KÕ ¡a§)� Â�� ¹¥ �a¨)§K�¢¡a§-¡¢¥�Ô�����ªÀ¨K�¢¡a��¥X¡¢¥ �a�K��Â

½��K¥��)�R��¥ ­¢¹ ¨)Ô �a�K�I«�� ��� ¡¢� ­¢¹2£ ¥�¤ �K��� «��)§K¡aÔ ¡a�)¤Ü¤ �)«��K¥�¤ �K¥�§K¹!���K­a¨K�¢¡a��¥��/Ñ�¨K½������)�K¥C¡a¤ �K¥ �¢¡aÔ ¡a�)¤C��¥C�¢Ñ����¨)� ¡a�s� Ô'� �¢��£�§K�¢£ ��¨K­�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥!��¥ ­¢¹�Ä��5���¢Ñ ¡a�7���)§)Â��¥�¤^� �a¨K¦��)Å��¢Ñ���­a�KÕ ¡a§)��¥U¡a�A¨)§)§)�)�����)¤s�a�s« ����½¡a¤ �-�KÕ�Â�¢��¨+§)��¥�¤�¡¢�¢¡a��¥��i��¥À«�¨K��� ¡¢¥ ¦�Ån�����¢Ñ�¨K���¢Ñ��^Ô ¡¢��� �5� �a¨K¦��«�¨K�����+§)¨K¥Ð���5¥���¥� ªÀ��¥����a��¥ ¡a§)¨K­¢­¢¹]¨K­¢�a�K���)¤�¡¢¥À½�¨K��¡aÂ��£��sÒ+¨K¹��U·����)�����)§K�¢¡a��¥°¿ Ä ¿�º�Ä���Ñ ¡a�U� �¢��¨K�a�K¦¹Xª�¡¢¥ ¡aª�¡a���)���¢Ñ��A¡¢ª�«�¨)§K�7� ÔM­a�KÕ ¡a§)¨K­�¦�¨K«��s· Ò5Ñ��K�¢Ñ��K�-¨K���¢Ñ��­a�K½��K­A� Ô�­a�Kª�ªÀ¨i���-� Ô��¢Ñ��i¨)����� §K¡a¨K�a�)¤^� £ ��§)¨K�a�K¦�����¡a¬�¨K�¢¡a��¥íÔ ��¨KªÀ�)� ºJ��¥4�¢Ñ��F� ¹�� �a�Kª «��K��Ô�����ªÀ¨K¥�§)�F· ¡¢¥«�¨K���¢¡a§K£ ­a¨K�A��¥i¡¢�a�'§)��½��K��¨K¦��Kº�Ä; <#= ô?>nõ ñ ô Í�Û @ ô�× Ë;"A î <#= ô < ô ÊÌË�B�Ø�Í�ñ Ú ÊDCE<#BGF��Ñ����a�)� �'§)����« £���§)��¥ �a¨K¡¢¥���¨/���K­a�)§K�¢¡a��¥À� Ôn���K¥ �a�K¥�§)�)��KÕ �¢��¨)§K�a�)¤^Ô ����ª¸�¢Ñ�����¨K­a¨K¥�§)�)¤/«�¨K���¢¡¢�¢¡a��¥ë� Ô��¢Ñ��'���a¨K­a¡a¨K¥à¶�¹¥ �a¨)§K�¢¡a§C¶��KªÀ¨K¥ �¢¡a§������)�K��¨K¥ áR· ��¶�¶��/ÅE¯°��¥�Â�a�KªÀ¨K¦¥ ¡Fü)ûÀú����H��´ ´ ´�º�Å-¡¢¥�§K­¢£�¤�¡¢¥ ¦(¨K���¢¡a§K­a�)��Ô ����ª

4JIDK7L)M'N'O P QSR'TUR'V�KWR'K�QSM'X,Y�P R'T?Z.[�K�P \]P N'R'O P [�V�^2V�M'_SR'O P [�V�^2`,V�[�V�aM'bSO X,R'c�[�d)M'K�efd)M'V�O M'V�N'MgR'X,_Sh�Z.M'V�O di`,d)h�Y7L)M'N'O j.[�Y7L)M'N'O j.P V�K�P X,M'N'O[�Y7L)M'N'O e'^%N'R'h�d)R'O P QSMkR'V�KlZ.[�K�R'TmN'[�V�d)O X,h�N'O P [�V�dS^%c�X,M'K�P N'R'O P QSMN'[�V�d)O X,h�N'O P [�V�dS^"nonpN'[�Z.c�T M'Z.M'V�O R'O P [�ViR'V�K&Z.[�K�P \]P N'R'O P [�V�^�M'Z.aY�M'K�K�M'KH\]P V�P O MqR'V�KHV�[�V�a,\]P V�P O MqN'T R'h�d)M'dS^oN'[�V�O X,[�Tr[�\DP V�\]P V�P O P QSR'Td)h�Y7L)M'N'O dS^sX,M'T R'O P QSM N'T R'h�d)M'dt`,Z.R'P VGN'R'd)M'd)e'^uc�R'X,O P N'P c�P R'ToN'[�V�d)O X,h�N'aO P [�V�dS^vR'K7L)M'N'O P QSR'TwN'[�[�X,K�P V�R'O P [�V�^vV�[�h�V�a,V�[�h�VxN'[�[�X,K�P V�R'O P [�V`,Z.R'P VyN'R'd)M'd)e'^�nonoa,nonzN'[�[�X,K�P V�R'O P [�Vp`,Z.R'P VpN'R'd)M'd)e'^�N'T P O P N'P {'R'aO P [�V�|

² å

Page 63: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

§)��¥ �a�Kª�«�����¨K��¹-���a¨K­¢¡a¨K¥v¥��KÒ+� «�¨K«��K����¨K¥�¤'«��K��¡a� ¤�¡a§)¨K­a�§)��½��K��¡¢¥ ¦ë¨�Ñ ¡¢¦Ñ+½�¨K��¡a�K�¢¹X� Ôp�a��« ¡a§)�7· «���­¢¡¢�¢¡a§)��Åp�)§)��¥�Â��ª�¹�ŧK£ ­¢�¢£ ���)Å��§K¡a�K¥�§)�)Å�Ñ��)¨K­¢�¢Ñ�Å%� «������aÅ�­a�K¡a� £ ���)Å%�K�a§)Ä º�Ä� © §)��¥�� ¡a� �a�s� Ôr��¿ Å ä ± ä5Ò+����¤5�a��á��K¥���Å�§)�������)� «���¥�¤ ¡¢¥ ¦@�a�}����±í���K¥ �a�K¥�§)�)�0· Ò5¡¢�¢Ñ�¨FªÀ�)¨K¥����K¥ �a�K¥�§)�­a�K¥ ¦�¢ÑX� Ô�¿ ¿ Ä ± åsÒ+����¤ ��Å�¡¢¥�§K­¢£�¤�¡¢¥ ¦�« £ ¥�§K�¢£�¨K�¢¡a��¥ç�a� Âá��K¥�� º�Ä���Ñ���ªÀ�)¨K¥0¥ £ ª����K�ë� Ô/¦��¨Kª�ªÀ¨K�¢¡a§)¨K­v���K­a¨)Â�¢¡a��¥���«��K�A���K¥ �a�K¥�§)�E¡a�'± å Ä;"A � <#= ô?~ Î�Ê ô�ó Û × ô �Î�Í�Ê ô Í?C ~ rF��Ñ��/��¨)���K­¢¡¢¥��/«�¨K�����K�7¡a�v¨s¥���¥� ­a�KÕ ¡a§)¨K­¢¡a���)¤i½��K��� ¡a��¥� ÔI�)�5»E�5��¡¢¥�§K­¢£�¤�¡¢¥ ¦:� �¢��£�§K�¢£ ��¨K­¢­¢¹� ��¨)���)¤2��£ ­a�)���¥ ­¢¹�Ä���Ñ��EªÀ�)¨K¥i¥ £ ª����K�A� Ô ¦��¨Kª�ªÀ¨K�¢¡a§)¨K­����K­a¨K�¢¡a��¥��«��K�A���K¥ �a�K¥�§)�s¤ �K�a�)§K�a�)¤�� ¹v��µ+¡¢¥i� © ¡a�'± ² Ä��Ñ��-��£ �¢« £ ��� Ô��¢Ñ��'��¨)���K­¢¡¢¥��'«�¨K�����K�E¡a��� Ñ�¨K­¢­a��Ò4¡¢¥¤�¡aÔ�Ô��K���K¥ �A���)� «��)§K�a��Är��¡¢��� �aÅ�¡¢�i§)��¥ �a¨K¡¢¥��7£ ¥�¤ �K��� «��)§K¡aÂÔ ¡a�)¤í¨K¥�¨K­¢¹����)��Å-���)�������a�)¤°�a�°Ò5Ñ��K¥��K½��K�4¨K½�¨K¡¢­a¨K� ­a�� �¢��£�§K�¢£ ��¨K­�¡¢¥�Ô�����ªÀ¨K�¢¡a��¥ë¤ � �)�n¥�����¨K­¢­a��ÒTÔ����/¨'ªÀ������ «��)§K¡aÔ ¡a§A� ¹¥ �a¨)§K�¢¡a§�¡¢¥ �a�K��« ���K�a¨K�¢¡a��¥�è ü�� ý�%¨K� �¢Ñ ¡a��­a�K½��K­aÅ¥��]¤�¡a� �¢¡¢¥�§K�¢¡a��¥F¡a�-ªÀ¨)¤ �U���K�¢Ò+�)�K¥O¨K��¦£ ªÀ�K¥ �a��¨K¥�¤ªÀ� ¤�¡aÔ ¡a�K����Å'Ò5Ñ ¡a§KÑ@¨K���F¨K­¢­�¦��K¥��K��¡a§)¨K­¢­¢¹I�a¨K¦¦��)¤I¨)�8)§)��ª�« ­a�KªÀ�K¥ �a�s:)Ä © ��¥�§)�K��¥ ¡¢¥ ¦[¨K�¢�a¨)§KÑ ªÀ�K¥ �aÅ��¢Ñ��7� ¹���Â�a�KªÝ�¢��¡a�)�^¨K­¢­ç� �¢��£�§K�¢£ ��¨K­¢­¢¹�Â�§)��ª�«�¨K�¢¡¢� ­a��¨K�¢�a¨)§KÑ ªÀ�K¥ �Ñ ¹«����¢Ñ��)���)�A¨K¥�¤s��¨K¥ á����¢Ñ��KªÜ¨)§)§)����¤�¡¢¥ ¦ç�a�U¨5§)��¥�Ô ¡a¤ �K¥�§)�Ý��§)�����)ÄI¶��¢����¥ ¦ « ���)Ô��K���K¥�§)�í¡a� ¦¡¢½��K¥ �a���¡¢¦Ñ �¢ªÀ� � �A¨K�¢�a¨)§KÑ ªÀ�K¥ �a��è�ü�� ý�M¨�« ���K«�� � ¡¢�¢¡a��¥�¨K­�§)��ªÀ« ­a�KªÀ�K¥ ��¡a�-¨K�¢�a¨)§KÑ��)¤iÒ5¡¢�¢Ñ��¢Ñ��sÑ ¡¢¦Ñ��)� �/§)��¥�Ô ¡a¤ �K¥�§)���§)������·�² ´�º/�a�U�¢Ñ���§Ì­K������e�KÅn���7��¡ ¦%Ñ� ª���e�KÅn¨Ì½M¨Ì¡ ­K¨Ì�­K�­K�ÌÕ¡K§�¨Ì­MÑ���¨�¤Ä)� ¥U� Ñ��-�̽M¨Ì­ £�¨Ì� ¡K��¥ç���Ì«������K��¤7¡ ¥Ð����§Ì� ¡K��¥��ÅnÒ^�X§���¥��e¡K¤�Ì���K��«� ��¨Ì¥áM��¤F¤�Ì«��Ì¥�¤�Ì¥�K����¥­ ¹MÅ%�,� ü��� Ñ�����^�Ì¥�Ô�����§Ì¡ ¥¦���¡ ¦%Ñ� ª���e�/¨Ì� �K¨�§ÌѪ��Ì¥�KÄ ¯I��������½M�Ì��Å¡ ¥�ª�¨Ì�K§ÌÑ¡ ¥¦]� Ñ��+���Ì­K¨Ì� ¡K��¥��i¹%¡K�Ì­K¤��¤Ð�¹]� Ñ��+«�¨Ì�����Ì�Ò/¡ � Ñi� Ñ��E��¶p¶��X���Ì­K¨Ì� ¡K��¥��p¡ ¥v� © Ò^�nª�¨ÌáM�7¨Ì­ ­K��Ò^¨Ì¥�§��Ô�������¥��M­K�̽M�Ì­��Ôp�e£���e£ªi«� ¡K��¥�Å��,� ü���¨���µi���Ì­K¨Ì� ¡K��¥U§�¨Ì¥������¥��-­K�̽M�Ì­�Ñ¡ ¦%Ñ��Ì�/� Ñ�¨Ì¥X¡ �K�A��¶p¶��@§���£¥�K�Ì��«�¨Ì���'¡ ¥� Ñ���Ñ¡K�Ì��¨Ì��§Ìѹ+�ÔM¤�Ì«��Ì¥�¤�Ì¥�§Ì¹/���Ì­K¨Ì� ¡K��¥���Äu�p¡ ¥�¨Ì­ ­ ¹MÅ�� Ñ����µJ��£� «£��¡K��«�¨Ì��� ¡K¨Ì­�Ò/¡ � Ñç�����e«���§Ì���K�7� Ñ�����/¤�Ì«��Ì¥�¤Â�Ì¥�§Ì¡K����·�ü�� ýD��¨+û ��úû¢Â §Ì­K¨Ì£��������^¨+¤�¡ ����§Ì�-���S����§Ì� ºs� Ñ�¨Ì�Ò+��£ ­a¤ç���5½��K��¹]¤�¡aÔ�Ô ¡a§K£ ­¢���a��¡a¤ �K¥ �¢¡aÔ ¹�Ò5¡¢�¢Ñ]¨^� £�Ô�Ô ¡a§K¡a�K¥ �5¤ �K¦���)�^� ÔA§)��¥�Ô ¡a¤ �K¥�§)�s�¢Ñ ����£ ¦ÑF� �¢��£�§K�¢£ ��¨K­¢­¢¹�Â��¨)���)¤���£ ­a�)�'��¥ ­¢¹�Ä;"A ; <#= ô?�nô�õ Û ��Î ó ó ð2� �H���"�i�S���E�S�i�r�"�S���S�?� � �H�r��.��v���'�������'�����D�)�'��� �w�'¡�¢���£W¤D�'¥*¦E��§�¡¨§�©Gª)«&¬ ­&®/��¡��'���£���¦/¯§�¢��°¦E¢�¥±��'¢���¥*�'�����D� ¯��¦s��£²�'¡£³���'�������'�����D�¯��¦s��£f¥±������¦%´ �¦E��¡��v¢���?µr­&¶m·?® ¬k���'������§�¡�¸*¹�ª)¡g¢�����¦���'�������'�����D�)�'��� �w�'¡�¢���£º��§�¡© ��� ��¥*�'¢���§�¡»rª)«&¬ ­&®¼© ��¥*¦E¢¢�¥±����¦?¢�§H��£��'¡�¢���© �½��¦?�w�'¡��½£��'¾�'¡£��'¡�'����¦&��¦.¾§�¦s¦E���¯����J¿&��¢��l¦E¢�¥±��'¢���¥*�'�?��¡©±§�¥±�w�'¢���§�¡¹�®f�'�������'�����D� ¯��¦s��£¥±������¦v��¡�¢��'¥±¤D�'¡�W���'¢��'¥y¢�§À¥*��© ��¡�½�'¡£�Á�§�¥À��§��v¾����'¢��¦E¢�¥±��'¢���¥*�'�����D� ¯��¦s��£Â�'¡�'���D¦s��¦s¹?�.�§�¦s��¦E¢�¥±��'¢���¥*�'�����D�

¯��¦s��£%��� ¾§�¢����¦s��¦r¢���'¢.© ��¡£m¦E��¾�¾§�¥±¢���¡H¢���#���'������§�¡�'¥*�Ã��¦s¦E��� ¡��£g¢���&����� ���¦E¢i¦s��§�¥*�ô±Ä�Å�¸*¹Æ�.��y��§�¡�¢�¥±���¯���¢���§�¡W§�©����'�������'�����D� ¯��¦s��£U¥±������¦���¦�ÇÈ�ÇÉ ÊÃÈ�ÇÈ�Ë�È�ÇÌ�Í"Χ���£Ã¥*�'���'¢���§�¡¦g���'¡Â�'¤D�'¡�¢���'�����À¯�3£�§�¿&¡�� ¥*��£���£�»?��¦¢���'�p��'¾�¾�'¡y¢�§p¦s��§�¥*��»���¡p¢���U¡�'¿&���g¥*�'¡�ÏD��£f����¦E¢G§�©¾§�¦s¦E��¯����¥*�'���'¢���§�¡¦s»m��§�¿Ã�'¥Ð¢���'¡¼¢���'��¥����'�������'�����D�¯��¦s��£y�'��¢��'¥±¡�'¢���¤D��¦s¹"Ñr��¥±¢���'¥±�w§�¥*��»2¦E¾���'��© �����'¢���§�¡k§�©��©±§�¥±�w�'¥Ã��¡£��'¥*¦E¾���'��© ����£À¥*�'���'¢���§�¡Ò��¦w�'��¿Ã�'�D¦w�������§��v¾�'¡�����£q¯��w�.Ó�Ô�É Ó�Õ�Ç�Ö�Ì Ç�×À§�©Æ¢���.¥*�'���'¢���§�¡¦t��£��'¡�¢���© ����£Ã©±§�¥i�%� ��¤D�'¡k¦s�'¡�¢��'¡����ØD© ¥*§��Ù¢�����¦#¥*��� ¥*�'¡�Ï ��¡��D»¥*��¦E¢�¥±��'¢���¥±��¡��3´±��¹ �D¹�¥*���'¢�¢����'���w�'¡�¢.§�©��§��v¾����'�w�'¡�¢�¦E¸§�©�¢���f© ��¡�'�.§���¢�¾���¢��w�'�3©±§�����§�¿Ã¹® ­&µº§���¢�¾���¢�¢����¦r��¡�'���£���¦sÎ�'¸Ú© �������k¦E¾���'��© ����£Ã£��'¾�'¡£��'¡�'�y¥*�'���'¢���§�¡¦sÎtÔ�Û ×Ûr�'¡

��¡£��'¥*¦E¾���'��© ����£k£��'¾�'¡£��'¡�'�À¥*�'���'¢���§�¡Ò¦E��'�Ò��¦Ü ��§��v¾����'�w�'¡�¢�ÝJ´,Þ2ßàin�¸*»*��£��'¡�¢���© ����£á¯��z�J¦E¢�¥±����¢���¥*�'�����D� ¯��¦s��£º¥±������»U��¦g¥*�'¿&¥±��¢�¢��'¡»?¿&��'¡Â���'��������'�����D�)¦E��¾�¾§�¥±¢���£�»%��¦ Ü ��¡£���¥*���'¢g§�¯�âo���'¢�Ýw´±·Gã2äåª)¸�'¡£m��¦s¦E��� ¡��£m� ����� ��'¥á��§�¡© ��£��'¡��� ¤D�'�����Ø

¯�¸æ¡�'¿²£��'¾�'¡£��'¡�'�º¥*�'���'¢���§�¡¦sÎ#¢�����¦J��¦J¢���w����¦s��»©±§�¥p��¡¦E¢��'¡����»f§�©vË çÕ�Ëå�)�'���'�¦s��¦s»q£���¥*���'¢Ã§�¯�âo���'¢�¦�'¡£g§�¢���'¥#¥*�'���'¢���§�¡v¢�� ¾��¦t¿&�§�¦s�.��£��'¡�¢���© �����'¢���§�¡��¦D¢��'ÏD�'¡g¢�§q¯�?¢�§�§g£���©±© ���'����¢U�'¡£q¡§���¦E�m¿&��¢��§���¢¦E��¾�¾§�¥±¢.§�©����'�������'�.�'¤ ��£��'¡����Ø

�'¸è��¡£��'¥*¦E¾���'��© ����£°£��'¾�'¡£��'¡�'�Ú¥*�'���'¢���§�¡¦s»é©±§�¥¢��§�¦s�f����¦s��¦r¢���'¢.© ��¡£%¡§%���'�������'�.¦E��¾�¾§�¥±¢�¹

�.��H�w���'¡k¡����v¯�'¥y§�©.� ¥*�'�v�w�'¢������'�G¥*�'���'¢���§�¡¦á¾�'¥¦s�'¡�¢��'¡���k£��'¢����'¢���£w¯��Ю ­&µ¨��¡l�Jê/��¦3ë�Ä�¹#ª)¡�¢,�D���¤t���,�D��¢,�'§S¡ì§ ©m¦u����¢,�'§S¡�íD»D¿y�w��§S¡D¦å�'£ ��¥H¢'§S¾D� ¥���¡ Ït��££ ��¾D��¡D£ ��¡ ¢'¦g§S¡ �,�Ò´7��§S¡D© �'£ ��¡D���3¦u��§S¥��

≥ î ÅS¸�»?��§S¥7¥����¦å¾D§S¡D£S�,¡ �q¢'§f���,¢,�D��¥��'��� �'�����,�,�t�E¦å� ¾ ¾D§S¥7¢'��£q£ ��¾D��¡D£ ��¡D���¥����'��¢,�'§S¡D¦m§S¥fïw�,¡w¢,�D���,¥g��¯D¦u��¡D���Jïw¢'§v¥7�,��� ¢,�z§ ¦å¢&��¢'�¢'����� �z��¡ ¢'¦u¹�­J�t���,¡D»��,¡ð�z��¢'��� �,¡ �ñ¢,�D�ò¥����'��¢,�'§S¡D¦���'���'£ ��£y¯ �À¢,�D�H¾D��¥�¦u��¥J¿J�,¢,�À¢,�D�iªsó2ó"�¨¥����'��¢,�'§S¡D¦á�,¡�fê3¿y� �z��Ït�f���,�'§S¿y��¡D���f©7§S¥á§S¡D� �'��¤t��� § ©D¦å� ¯D¦å� �m¾D�¢,�'§S¡D¹ôõ�H���"ö ÷2��ø �Dù"ú�ûG�S����ö �E��.��J¾�'¥*¦E��¡��Ч���¢�¾���¢�¦H§�©#ü%µ¨�'¡£m® ­&µº¿Ã�'¥*�p��§��w�¾�'¥*��£k�'¡£y¾�¥*§7âo���'¢���£À�'�D�'��¡¦E¢GªSóróo�}�'¡�¡§�¢��'¢���§�¡À¢�§��¦s¦s��¦s¦?¢���g��§�¡�¢�¥±��¯���¢���§�¡�§�©D���'�������'�#��¡©±§�¥±�w�'¢���§�¡W¢�§¾�'¥*¦s�l¦E�������¦s¦s¹Gª)¡¨¢�����¦y¾�'¾�'¥*»U¿Ã�Щ±§��'�¦À§�¡ý¢����'¤D�'����'¢���§�¡/§�©á�§�¿þ�'¡£z¢�§3¿&�����'�ì�'��¢��'¡�¢f���'������§��¦E� ¡�¢����'¢����z��¡©±§�¥±�w�'¢���§�¡ÿ��§�¡�¢�¥±��¯���¢���¦p¢�§k��£��'¡�¢���© �������¢���§�¡ÿ§�©i¢���z¾�¥*§�¾�'¥º�'¢�¢����'���w�'¡�¢z§�©&¾�¥*�'¾§�¦E��¢���§�¡�'���§��v¾����'�w�'¡�¢�¦s¹�ÑU§�¥i�'¡º��¦s¦s��¦s¦E�w�'¡�¢á§�©r¢���%¥*§����m�'¡£���v¾���'¢H§�© ���'�������'�U��¡©±§�¥±�w�'¢���§�¡���¡�¢���Ã�'¡�'���D¦E��¦m§�©£��'¾�'¡£��'¡�'�W¾�'��¥*¦?�����£���£H¯��½¦E¾���'��© ���q¿Ã§�¥*£�¦s» ¢�����¡�¢��'¥*��¦E¢���£À¥*����£��'¥y��¦v¥*��©±�'¥±¥*��£À¢�§�üm�'¥±¢�§�����¡��ÃÔ�ËiÕ��å¹´ �Å�Å���¸*¹

�

Page 64: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

ô�� � �f���"���Eø �E�"�Eø 2���2�"ö ���"�Eø ù"��J�'¯����WëÀ¦E���v�w�'¥±��¦s��¦%¢���&¥*��¦E����¢�¦m§�¯�¢��'��¡��£g¯��w¢���¢�¿Ã§}£���©±©±�'¥*�'¡�¢�¾�'¥*¦E��¡�� ��§�¡© ��� ��¥*�'¢���§�¡¦Ò´ ü%µ²�'¡£® ­&µr¸Ù§�¡æ¢���¨¢���¦EÏ §�©Ò�'¢�¢����'����¡��ò¾�¥*�'¾§�¦E��¢���§�¡�'���§��v¾����'�w�'¡�¢�¦º´ µUêD¸*¹%µr¥*�'¾§�¦E��¢���§�¡�'����§��v¾����'�w�'¡�¢�¦�'¥*�f�'����¦s¦E��© ����£G¿&��¢��H¥*��¦E¾���'¢Æ¢�§G¢���#�D§�¤D�'¥±¡���¡��i�����£�εUê� ����&­ ¥*��©±�'¥*¦Ð¢�§³�'���½¾�¥*�'¾§�¦E��¢���§�¡�'�¨��§��v¾�������w�'¡�¢�¦}�D§�¤D�'¥±¡��£è¯����W´±�'¥±¯�'��¸*»��ô±§��v��¡�'��¸è§�¥­Ã´±£7âo���'¢���¤D�'��¸p�����£�¦s¹ µUê� �� ��¦m¢���k¦E��¯¦s�'¢&¿&��¢��x��W´±�'¥±¯�'��¸Â�����£æ�'¡£ µUê� �� ¢���}¦E��¯¦s�'¢�¿&��¢�� ��ô±§��v��¡�'��¸p�����£�¹ ÑU§�¥À�����'�µUêò�'����¦s¦s» ¾�¥*���'��¦E��§�¡»¥*�����'���.�'¡£m©¦s��§�¥*�f© ��� ��¥*��¦U�'¥*� � ��¤D�'¡z©±§�¥¢���f£���©±©±�'¥*�

�'¡�¢á¾�'¥*¦E��¡��Â��§�¡© ��� ��¥*�'¢���§�¡¦s¹2µr¥*���'��¦E��§�¡k��¦g£���© ��¡��£��¦m¢���y¥*�'¢���§l§�©g��§�¥±¥*���'¢����Ð��£��'¡�¢���© ����£l£��'¾�'¡£��'¡�'�¥*�'���'¢���§�¡¦Ã§�¤D�'¥z�'���q¥*�'���'¢���§�¡¦y©±§���¡£W¯��º¢���v¾�'¥*¦s�'¥´ ¾�¥*�����Ð��§�¥±¥*���'¢����v��£��'¡�¢���© ����£q¥*�'���'¢���§�¡¦DÁr¢�§�¢��'�2¡����w�¯�'¥%§�© ��£��'¡�¢���© ����£G¥*�'���'¢���§�¡¦E¸*Ø*¥*�����'����¥*��©±�'¥*¦�¢�§U¢���¥*���¢���§½§�©H��§�¥±¥*���'¢����º��£��'¡�¢���© ����£½£��'¾�'¡£��'¡�'�k¥*�'���'¢���§�¡¦§�¤D�'¥Â�'���3¥*�'���'¢���§�¡¦w��¡ ªSóróo�ð´ ¥*�����'�����è��§�¥±¥*���'¢������£��'¡�¢���© ����£á¥*�'���'¢���§�¡¦tÁ2ªSóróo�k¥*�'���'¢���§�¡¦E¸*¹�Ñr��¡�'�����D»�¢���§�¤D�'¥*�'���g¾�'¥*©±§�¥±�w�'¡���l§�©i¢���z¾�'¥*¦E��¡��Ù¦E�D¦E¢��'�w¦p��¦£���¦s�'¥±��¯��£W��¡½¢��'¥±�w¦y§�©G¢���z©i¦s��§�¥*��»U��§��v¾���¢���£º��¦©±§�����§�¿Ã¦sÎ��m¾�¥*��� ¥*�����'����Á�¾�¥*�����W¥*�����'����¹

��� ����� "!�!�# n%$'&)( $'&)()*)+ + ,%d"()-�$'& n%$'&)( $'&)()*)+ + .od"()-�$'&��/�0�1�2�� 3�4�5�6 7�5 j 5�3 5�7 j 4�8 9�5 j : 3 7�4 j 6 : 9�; j 8 : 9�7 j :�:��/�0�1 ;<5�3 : 7�5 j 4�3 4�5 j 5�8 5�9 j 7�9 7�4 j : 3 4�= j 5�8 9�; j :�:��/�0�2 ;<6�3�5 7�3 j 5�3 6�8 j 6 : 7�7 j 8�8 7 :�j 7�9 6�; j 3�9 7�9 j 6 :

�J�'¯����fë�¹�µr¥*�'¾§�¦E��¢���§�¡�'�.��§��v¾����'�w�'¡�¢.�'¢�¢����'���w�'¡�¢���¡mü%µº�'¡£%® ­&µ

�J�'¯������¹�®f�'�������'����¦s��£m�'¢�¢����'���w�'¡�¢�¦

�J§g©±§��'�¦G§�¡v¢���.¥*§����J§�©�¢���.���'������§�¡v��¡w�'��¢���'¥q��§�¡�© ��¥±�v��¡�� §�¥½¥*�'¤ ��¦E��¡�� ¦E¢�¥±��'¢���¥*��� ¯��¦s��£Ù£��'¾�'¡£��'¡��'����¦s»����'�������'�����D�)¦E��¾�¾§�¥±¢���£p�'¢�¢����'���w�'¡�¢�¦á�'¥*�&¦E��¡�� ����£§���¢i©±§�¥p�'¤D�'����'¢���§�¡���¡��J�'¯����>�¹��.��'��¥g�'���v�����'¢���¤D�© ¥*��?���'¡�'�Â��§���¡�¢�¦p�'¥*�m¥*�'¾§�¥±¢���£Ã��¡º¢���3© ��¥*¦E¢%¢���¥*�����§������v¡¦%§�©Æ�J�'¯����@�w´ Ü ®f�'�������'����¦s��£v�'¢�¢����'���w�'¡�¢�¦sÝ'¸*»¢�§��D�'¢���'¥%¿&��¢��3¢���'��¥H£���¦E¢�¥±��¯���¢���§�¡W¾�'¥G�����£W���'¢��'�D§��¥±����¦s¹ ®f�'�������'����¦s��£W�'¢�¢����'���w�'¡�¢�¦.��¡�'���£��ᯧ�¢��W¢��§�¦s�¦E¢�¥±��'¢���¥*��� ¯��¦s��£/�'¢�¢����'���w�'¡�¢�¦W¢���'¢y��'¾�¾�'¡ì¢�§�¯���§�¡© ��¥±�w��£g���'�������'�����l´ Ü ê%§�¡© ��¥±�w��£��'¢�¢����'���w�'¡�¢�¦sÝ'¸*»�'¡£i¥*��¦E¢�¥±��'¢���¥*��£W�'¢�¢����'���w�'¡�¢�¦s»2Ì�Û Ô�Û�¿&��'¡º�%¾�¥*�'¾§��¦E��¢���§�¡�'�y��§��v¾����'�w�'¡�¢J¾�¥*�'¤ ��§��¦E���ì�'¢�¢����'���£3¢�§3¢����'��§�¦s��¦E¢%�'¤D�'�����'¯����G�����£J¢�§f��¢�¦����© ¢t��¦á�'¤D�'¡�¢���'�����g¥*�����¦s¦E��� ¡��£���¦%¢���ã��'¾�'¡£��'¡�¢i§�©q�y©±�'¥±¢���'¥q�����£�»#§�¡¢���p¯��¦E��¦Ã§�©%���'������§�¡���§�§�ÏD� ��¾ì´ Ü ¶m��¦E¢�¥±��'¢���¥*��£��'¢��¢����'���w�'¡�¢�¦sÝ'¸*¹r�J�'¯����A�º¢����¦p¦E�§�¿Ã¦f¢���m���v¾���'¢m§�©���'�������'����¡©±§�¥±�w�'¢���§�¡z§�¡m¢��� ¢���¦EÏ3§�© µrµk�'¢�¢����'���w�'¡�¢�¹ª)¡y�w§�¦E¢%����¦s��¦s»CB ��D §�©�¢���U¢�§�¢��'�G§�©����'�������'����¦s��£p�'¢��¢����'���w�'¡�¢�¦s»2® ­&µÐ¯��¦E�����'�������§�¡© ��¥±�w¦v£��'¾�'¡£��'¡�'�¥*�'���'¢���§�¡¦i�'��¥*����£�����¦s¦E��� ¡��£z�'¢.¢���q¾�¥*�'¤ ��§��¦&¦E¢��'�D��¹

�Ã�'¿&��� £���¦s��§�¤D�'¥*��£x�'¢�¢����'���w�'¡�¢�¦s»q¿&�����'�ò�'¥*�£����¢����'¢���£i¢���'¡�ÏD¦.¢�§i���'������§�¡W��§�§�ÏD� ��¾½�'¡£i¥*��� ¥*�'¡�Ï ��¡��D»�'�w§���¡�¢v¢�§ý§�¡����Ùë�ë D §�©Ã�'���m���'�������'����¦s��£Â�'¢�¢����'���w�'¡�¢�¦s»3����¦s¦Â¢���'¡�E D §�©ì�'���lµrµ°�'¢�¢����'���w�'¡�¢�¦� ���'��£���£%¯��H® ­&µU¹ô�� F GGø ���S������ø ù"�H�I J�I KMLON�P�Q�RSR>Q�T�UWVYXZN�P�[S\Z[S]�T^]�T^T�]�_�T^Q�T�U�`�N�XZab�N�Q�U�\®f�'¢t�¦á��§�¡¦E��£��'¥.¢���i§���¢�¾���¢G§�©�ü%µÐ© ��¥*¦E¢�¹��.��i¦E¢�¥±��ÏD���¡��é£���©±©±�'¥*�'¡���W��¡¢���Ã¥*�����'���W§�©f¡§���¡� �����£���£dc�e¤D�'¥±¯� �����£���£i¾�¥*�'¾§�¦E��¢���§�¡�'�q�'¢�¢����'���w�'¡�¢�¦J´±§�¡º��§��w�¾�'¥*�'¯����q���'¤D�'��¦&§�©D¾�¥*���'��¦E��§�¡» ¥*§�¿Ã¦��½�'¡£�EW§�©t�J���¯�����ë�¸Ã¾�¥*§��v¾�¢�¦g¢����¦E��� �D��¦E¢���§�¡Â¢���'¢m¢���W¢�� ¾������'���§�¡�¢��'��¢.§�©��¦s�f§�©� ¡§���¡m��¦r�w§�¥*�f����¦E�����3£���¦s�'¥±��¯��£��¡À¢��'¥±�w¦v§�©.��§����'��».§�¥*£��'¥*�)��§�¡�¢���¡��D�'¡�¢H�'¥±��¢��'¥±���W´7Ô�Û ×Û¥±��� ��¢��w§�¦E¢?�'¢�¢����'���w�'¡�¢�¸#¢���'¡w�.¤D�'¥±¯z��§�¡�¢��'��¢���¦s¹�fé����'¡¨� ��¤D�Ð�'¢v������¦E¢v¢���¥*���z¥*����¦s§�¡¦À©±§�¥W¢���'¢�¹UÑr��¥*¦E¢�»© ¥*�'�w�m¯���'¥±��¡��À¡§���¡¦q¢��'¡£y¢�§k¦s�'�����'¢m©±�'¿Ã�'¥Ã�'¥±� ��

��g)hji k)lnm i o"g)pqlnr'r'o /�snt�u'i v)wqg)pqlnr'r'o x�g)o"r'v)y�k)r'y�v)g)pqlnr'r'oz - z *)+ ß�{ c�$'&)( |}- z *)+èß�{ c�$'&)( z - z *)+ ß�{ c�$'&)(

����0�1�2�� =�;<= 6�;<= 6�= j ; : 6�;<9~7�7�;�=�4 j 4�= ;<8�3 9�5 9�3 j ;�;����0�1 : 6�= : 4�4 6�4 j 4�3 : 8�;~;<=�4�=�9 j 5 : 6�6 9�; 9�= j 3 :����0�2 9 : = 5�7�5 =�; j 4�; 9�;<4~5�7�7�=�3 j =�7 ;<5 4 : 9 j 9�7

Ä�Å

Page 65: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

�w�'¡�¢�¦?¢���'¡3¤D�'¥±¯¦&£�§�¹ ª)¡�§���¥%���'������§�¡»Dë�Ä � Ei¤D�'¥±¯������£���£�© ¥*�'�w��¦H§���¢&§�©qÄ � ��íw��'¤D�J�w§�¥*�J¢���'¡l§�¡�¡§�¡z¦E��¯�âo���'¢.�'¥±� ���w�'¡�¢.´ ��í¹ í D ¸*»�¿&�������#¢���'¥*�#¯�'��¡��§�¡����æë ��î Å¡§���¡� �����£���£Ù© ¥*�'�w��¦§���¢�§�©kë î E ���¿&��¢��g�w§�¥*�?¢���'¡w§�¡�J�'¥±� ���w�'¡�¢?´±ë��¹ Ä D ¸*¹*ª)¡v�Jê%»�§�©��E�Å�Åw�����£z¤D�'¥±¯Ð¢�§�ÏD�'¡¦s»YE���BÐ�'������¯���¢J�w§�¥*�p¢���'¡§�¡�J¡§�¡l¦E��¯�âo���'¢J�'¥±� ���w�'¡�¢J´±ë�í D ¸*¹�¶%��� ��¢��w§�¦E¢J�'¢��¢����'���w�'¡�¢ ¢�¥±��¤ ���'����� ¾�'¡�'����¦s��¦ð¦E��'� �'¥±� ���w�'¡�¢�'��'��¡¦s»H¿&��'¥*�x¦s§��w�x�'¥±� ���w�'¡�¢�¦k��'¾�¾�'¡ÿ¢�§ý¯�§�¤D�'¥±¢����y¥*���'����¦s��£&��¡º��§�¡�¢��'��¢á§�¡�m§�¥U�w§�¥*�m¦E¢��'¾¦ ¥*����w§�¤D��£Ð© ¥*§��è¢���'��¥m�����£�¦s¹D�.���¦s����§�¡£3¥*����¦s§�¡���¦¦s�'¡¦E��¢���¤D�H¢�§y���'¡�� ��'�D�H¤D�'¥±���'¢���§�¡ÎD¤D�'¥±¯Â�'¥±� ���w�'¡�¢�¦¢��'¡£g¢�§g¯�ã���¦E��§����'¢���£g�w§�¥*�y����¦E�����z¢���'¡w¡§���¡l�'¥*�� ���w�'¡�¢�¦s»r��¦f£���¦E��§����'¢���§�¡Ã����'¤ �����k£��'¾�'¡£�¦f§�¡º¦s�'¡�¢��'¡����� ���'¤D�'�v´ ��'¡���m�w�'��¡º¤D�'¥±¯� ���'¤D�'��¸&¾���'¡§��w�'¡�¦E��'�ÿ��¦À¦E����© ¢w§�©i¢�§�¾����Ч�¥½�'�v¾����¦E��¦s¹Uª)¡ýª)¢��'�����'¡»¢�§�¾������)£�¥±��¤D�'¡}�'¥±� ���w�'¡�¢�£���¦E��§����'¢���§�¡¨��¡ý¾�¥*�'¤D�'¥±¯�'�¾§�¦E��¢���§�¡Ù��¦��§��v¾�'¥*�'¢���¤D�'���æ© ¥*��?���'¡�¢Ð�'¡£Â¥*�'¾�¥*���¦s�'¡�¢�¦W�p¾�¥*§�¯����'�Ú©±§�¥v¢���g¯��¦s�'����¡�g¾�'¥*¦s�'¥*»D¿&�����'�¿Ã§�¥±ÏD¦%§�¡��?�����£��)© ��¥*¦E¢U��¦s¦E���v¾�¢���§�¡¹��.����¥*£����D»*¤D�'¥±¯¦�'¥*� ¢�� ¾������'�����H�w§�£���© ����£%¯��3� ¿&��£��'¥á¦s�'¢.§�©D¦E� ¡�¢����'¢����¦s�'¢��'������¢���¦J¢���'¡½¡§���¡¦p�'¥*��»U¦E��'�ý��¦f¢��'�v¾§�¥*�'�v�'¡£�'��¥*�'���w¦E¢��'¡�¢����'�U�w§�£���© ���'¥*¦m´ «&��Ïlë � B � ¸*¹�ÑU§�¥g�'��'�w�¾�����»G£��'¤D�'¥±¯�'�q¡§���¡¦y£�§W¡§�¢q��¡���'¥±��¢q¢���v¾§�¦s¦E��¯����¢��'�v¾§�¥*�'�.�w§�£���© ���'¥*¦i§�©¢���'��¥%¤D�'¥±¯3¯��¦s�g´ �%Ó���Ç�Ë çÔÊÃÕ�Ó�Õ�Ë çÈ�Ç}Ì Ç}Ë ç�Ó�Ô�ÔwçÈ���ÓZe »G¯���¢�� Ë çÔzÓ���ÇÿÈ �xË çÔÊÃÕ�Ó�Õ�Ë çÈ�Ç½Ì Ç½Ë ç�Ó�Ô�ÔáçÈ���ÓZeS¸*¹��ý§�£���© ���'¥*¦J§�©r¢�����¦f¦s§�¥±¢¢��'¡£á¢�§á¯�J£���¦E¢�¥±��¯���¢���£á��¡v¢���J¦s�'¡�¢��'¡���.�v��'�v�w§�¥*�© ¥*���'���H¢���'¡z§�¥*£���¡�'¥±�3�'¥±� ���w�'¡�¢�¦s¹H�I J�I �����@VYQ�P��C]<�W�Sb�N�RSN)��[SP�]�T�]�T�XZN�P�Q�RSR·&©U¢���m¢���¥*���3�'¯§�¤D�H�w�'¡�¢���§�¡��£k©±���'¢�§�¥*¦s»?§�¡����À¢���© ��¥*¦E¢Ã§�¡�y���¦3�'¡x§�¯�¤ ��§��¦H���'�������'�Ã�'��'¥*���'¢��'¥*¹�fé����'¡Ð¾�¥*§�¤ ��£��À�p¥*§���� �/��¦E¢����w�'¢��À§�©á¢���p���v¾���'¢y§�©���'�������'� ��¡©±§�¥±�w�'¢���§�¡½§�¡W¢���á¾�'¥*©±§�¥±�w�'¡���v§�©t® ­&µU¹�.��m���'������§�¡ý© ����¢��'¥W��§�¡�¢�¥±��¯���¢���¦g� ��D ��¡�'¥*����¦s�3§�©¥*�����'���W§�¡l¤D�'¥±¯x��§��v¾����'�w�'¡�¢�¦3´ í D §�¤D�'¥gí î ¹ î�D ¸*»¯��À��§�¥±¥*���'¢����p¥*���'¢�¢����'����¡��p¢�§J¢���G¤D�'¥±¯�'�t�����£f¢��§�¦s��'¥±� ���w�'¡�¢�¦À´±Ä�ë�¸W¢���'¢v¿Ã�'¥*�3¿&¥*§�¡�� ��� �'¢�¢����'���£k¢�§¢���'��¥i���v�w��£����'¢��'���k¾�¥*������£���¡��Ò��§�¡¦E¢���¢���'¡�¢%¯��Àü%µU¹�.����¦g������£�¦g¢�§Â�'¡Ù§�¤D�'¥*�'���mí � ¹ î�D ¥*�����'����¹?­&���m¥*����w�'��¡���¡��3©±�'��¦s� ¡�'�D�'¢���¤D��¦U´±�'¯§���¢�í�B D ¸á�'¥*� ��¸á�'��¢���'¥¤D�'¥±¯À�w§�£���© ���'¥*¦v§�¥J����¸J¾�¥*§�¾�'¥J¤D�'¥±¯Ò�'¥±� ���w�'¡�¢�¦á���D���¡��§���¢v§�©U¢���m¥*�����'�§�©&¦E¢�¥±��'¢���¥*��� ¯��¦s��£k�'¥±��¢��'¥±����»£���?¢�§v¦E� ¡�¢����'¢����.¾���'¡§��w�'¡�J¦E��'�w��¦G��§��v¾����'�w�'¡�¢£���¦E��§����'¢���§�¡»W��§��v¾����'� ��§�§�¥*£���¡�'¢���§�¡»&¾�'¥*�'¡�¢���'¢������§�¡¦E¢�¥±��'¢���§�¡¦p�'¡£º�'������¾¦E��¦s¹�fé�z¦E��'���á¥*�'¢���¥±¡º¢�§k��w§�¥*�g£��'¢��'������£W�'¡�'���D¦E��¦J§�©G©±�'��¦s�á¡�'�D�'¢���¤D��¦.��¡½¦s�����¢���§�¡pí¹ E�¹ í¹�ª)¡p¢���i����¦s�i§�©�¡§���¡���§��v¾����'�w�'¡�¢�¦s»��¦s�§�© ���'�������'�U��¡©±§�¥±�w�'¢���§�¡�¾�¥*§�£������¦m�&¡�'� ����� ��¯����&��¡��'¥*����¦s�v§�©r¥*�����'����ÎÅ�¹ Ä D ´UÅ�¹ î�D §�¤D�'¥�B�Å�¹ B D ¸*¹o�.����¦

��¦r¡§�¢.¦E��¥±¾�¥±��¦E��¡��D»���¦U§���¥¢���¦E¢.��§�¥±¾��¦U��§�¡�¢��'��¡¦2¤D�'¥±�©±�'¿ ����¦s��¦ÿ§�©Ð¡§���¡� �����£���£ �'¥±� ���w�'¡�¢ �'��'��¡¦s»©±�'¿Ã�'¥G¢���'¡W¿Ã�m��§�����£Ã�'��¾���'¢��©r¢���%¾�¥*§�¯�'¯�������¢��k§�©¢���'��¥á§����'��¥±¥*�'¡��� ¥*��© �����'¢���£%¢���q´ ��¡���©±§�¥±�v¸D¢�� ¾�q£���¦s�¢�¥±��¯���¢���§�¡�§�©Æ¡§���¡�© ¥*�'�w��¦D��¡g¢���?���'������§�¡¹*�.��.¤D��¦E¢�w�,âo§�¥±��¢��Ò§�©?¡§���¡� �����£���£k©±�'��¦s�H¡�'�D�'¢���¤D��¦s».��¦á¿Ã�¦E��'���g¦s���m��¡º�w§�¥*�z£��'¢��'���á��¡ý�m�w§��w�'¡�¢�»r��¦f¥*�'¾�¥*���¦s�'¡�¢���£%¯��H�w§�£���© ���'¥*¦s¹H�I J�I J����@VYQ�P��C]<�W�Sb�N�RSN)��[SP�]�T�]�T�VYXZN�P�[S\Z[S]�T¶m���'¢�¢����'���w�'¡�¢º��¦ì�'¡©±§�¥*����£x¯��¼® ­&µ ¿&��'¡ ¢���¾�¥*�'¾§�¦E��¢���§�¡Ã��¡�¢�¥*§�£���'��¡��k�m���'¡£���£��'¢��H��§��v¾����'�w�'¡�¢��¡é��§�¡�¢��'��¢i��¦z©±§���¡£���¡¢���Ã���'�������'�W© ¥*�'�w�º§�©q��¢�¦�����£�¹t�J�'¯����W�¨¦E�§�¿Ã¦J¢���'¢���E�� D §�©U¢���zë�Å�EÃ¥*���¦E¢�¥±��'¢���¥*��£º�'¢�¢����'���w�'¡�¢�¦q¾�¥*§�¾§�¦s��£y¯��À¢���H���'������§�¡�'¥*�ÿ&¥*§�¡��D¹ ¬ ¤D�'¡l�w§�¥*�y��¡�¢��'¥*��¦E¢���¡�� ���D»#¢���'¥*�y��¦3�¦E¢�¥*§�¡�� ��¦E� �v�w�'¢�¥±�ò¯�'¢�¿Ã���'¡³¡§���¡¦ì�'¡£x¤D�'¥±¯¦s¹f��¢��¨¤D�'¥±¯ý�����£�¦s»U¾�¥*���'��¦E��§�¡ÿ§�©i���'�������'�����D�)£�¥±��¤D�'¡¥*���'¢�¢����'���w�'¡�¢�¦��¦á©±�'��¥±���g����� ��´ ����Å D ¸*»�¡§�¡�'¢���'����¦s¦��§�¿Ã�'¥½¢���'¡ÿ¾�¥*���'��¦E��§�¡ §�©W¥±��� ��¢��w§�¦E¢Ð�'¢�¢����'���w�'¡�¢´ ��� î�D ¸*¹Æª)¡�¢���Ã����¦s�ç�© ¡§���¡������£�¦s»�¢���J¡����v¯�'¥§�©f���'�������'�����l¥*���'¢�¢����'���££��'¾�'¡£��'¡�'����¦m��¦H��¡¦E¢�����£�'��¢�¥*�'�w�'���ý��§�¿Ã¹G�.��w¾�'¥*���'¡�¢��'�D�l§�©H�v��¦E¢��'ÏD��¦p��¦����� �»�¿&��¢��m¾�¥*���'��¦E��§�¡z£�¥*§�¾�¾���¡��H¢�§O�Ä�¹ Ä D ¹�.��y£���©±©±�'¥*�'¡���f��¡z¢���f¢�§�¢��'�.¡����v¯�'¥v§�©¥*��¦E¢�¥±����¢���¥*��£W�'¢�¢����'���w�'¡�¢#�w�'�ï�v�'�D�'��¡½£���á¢�§&¢���%¥±���'��'¥��§��v¾����'�w�'¡�¢��'¢���§�¡�¾�'¢�¢��'¥±¡¦m�'������¯���¢���£g¯��z¤D�'¥±¯¦G��¡¢���U���'������§�¡¹��ç�¿Ã�'¤D�'¥*»�¿&�������U��¡p¢���i����¦s�i§�©�¤D�'¥±¯¦���'�������'�p��¡©±§�¥±�w�'¢���§�¡ì¾�¥*§�£������¦º�¦E��� ¡���© �����'¡�¢g���w�¾�¥*§�¤D�'�w�'¡�¢%§�¡y¥*��¦E¢�¥±��'¢���¥*��£y�'¢�¢����'���w�'¡�¢D¾�¥*���'��¦E��§�¡»¢�����¦Ã��§�¡�¢�¥±��¯���¢���§�¡ì£�¥*§�¾¦y��§�¡¦E��£��'¥*�'¯����ý©±§�¥H¡§���¡¦s¹�.��J�w�'��¡w¥*����¦s§�¡Ð©±§�¥á¢�����¦i¦E��¢���'¢���§�¡z��¦U¢���'¢.¡§���¡¦¢��'¡£&¢�§Ã¦s�'�����'¢á¦s�'�w�'¡�¢������'�����y¤D���'�§��¦ ¾�¥*�'¾§�¦E��¢���§�¡¦¦E��'�z��¦?È �&�v��'�m�w§�¥*�f§�© ¢��'¡m¢���'¡m¤D�'¥±¯¦U£�§�¹�ª)¡z§���¥���'������§�¡»?§���¢m§�©?íë î �k© ¥*�'�w��¦á�����£���£p¯���i¡§���¡»íÅ�ë î ��§�¡�¢��'��¡À¢���&¾�¥*�'¾§�¦E��¢���§�¡d�ÌG��¦m�'¡�'¥±� ���w�'¡�¢��¡�¢�¥*§�£�����'¥�´ � Ä�¹ Ä D ¸*¹��ÃÌU��¦H��¡x©±���'¢y�'¡/�'��¢�¥*�'�w�'���¾§����D¦s�'�w§��¦ ¾�¥*�'¾§�¦E��¢���§�¡»o�����£���¡��D»r�'�w§�¡��k§�¢���'¥*¦s»�'��¦s§q¾§�¦s¦s��¦s¦E��¤D�.¾���¥*��¦s��¦G�'¡£v§�¢���'¥#Ï ��¡£�¦G§�©��w§�£����© ���'¥*¦s¹S�.����¦2¢�¥±��¤ ���'�����i��¡�'¥*����¦s��¦2¢���¡����v¯�'¥G§�©t����¦s��¦§�©?�'¢�¢����'���w�'¡�¢%�'�v¯���� ����¢��À�'¡£y�'¤D�'¡�¢���'�����p¢���G¾§�¦s�¦E��¯�������¢��k§�©r�D�'¢�¢���¡��k©±�'��¦s�%¾§�¦E��¢���¤D��¦s¹rê%§�¡�¤D�'¥*¦s�'���D»2��¦¦E�§�¿&¡z¯��W¢���q¡����v¯�'¥m§�©%��§�¡© ��¥±�w��£3�'¢�¢����'���w�'¡�¢�¦��¡g�J�'¯������»�¢���?¥*§����&§�©Æ���'�������'�r��¡©±§�¥±�w�'¢���§�¡v��¡w© ��¥*�¢���'¥q¦E¾���'��© � ��¡��z�'¡w�'¢�¢����'���w�'¡�¢2¿&��¢��m¡§%¥*��¦E¢�¥±��'¢���¥*���¡��H��¦U�'���w§�¦E¢���¡���©±§�¥±� ���'¥*§�¦s¦r¡§���¡¦U�'¡£%¤D�'¥±¯¦s¹H�I J�I HM�OQ�RS\ZN�T�N)��Q��S[S`�N�\�.��3¤D��¦E¢m�w�,âo§�¥±��¢�� §�©&��¡£��'¢����'¢���£k¤D�'¥±¯ ��§��v¾�������w�'¡�¢�¦½´�B�Å�¹ Ä D ¸l�'¥*�À�w§�£���© ���'¥*¦½§�©v¤D�'¥±��§��¦ÃÏ ��¡£�¹�.��f¥*�'�w�'��¡���¡���¦s�'¢f§�©%©±�'��¦s�q¡�'�D�'¢���¤D��¦&��§�¡¦E��¦E¢�¦&§�©

Ä�ë

Page 66: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

í�B ��§��v¾����'�w�'¡�¢�¦Ù´���¹ � D ¸*»�E�Å}��¡£���¥*���'¢x§�¯�âo���'¢�¦´ í¹ B D ¸p�'¡£gí�E�§�¯�����?���y�'¥±� ���w�'¡�¢�¦H´±Ä�¹ ��D ¸*¹��ý§�¦E¢¦E��'�þ��§��v¾����'�w�'¡�¢�¦é�'¥*���<�?� ¾���¥*��¦s��¦l��¡æ¾��¦s¦E��¤D���§�¡¦E¢�¥±��'¢���§�¡¦J¿&�����'�¨�'¥*�m¡§�¢v��¦p¦E��'�º¤D�'¥±�£���©±© ����'����¢%¢�§k£��'¢����'¢%¯���¢Dâu�¦E¢%��'¾�¾�'¡À¢�§À©±�'���H§���¢H§�©.¢����'��¥±¥*�'¡�¢ ��§�¤D�'¥*�'�D�q§�© ® ­&µU¹n�ý§�¥*�#��¡�¢��'¥*��¦E¢���¡�� ���D»n��Á�E§�© ¢���&¥*�'�w�'��¡���¡��l©±�'��¦s�J¡�'�D�'¢���¤D��¦H�'���£��J® ­&µ½¯������'�¦s�z¢���'� �'¥*��§�¤D�'¥±¢����Ò¥*���'����¦s��£Â©±�'¥º�'¿Ã�'� © ¥*§��¢���'��¥i¤D�'¥±¯k�����£�»?§�© ¢��'¡k¢�§y��¢�¦q����© ¢�¹��ý§�¦E¢m§�©?¢����¦s���§�¡¦E¢�¥±��'¢���§�¡¦J��¡�¤D§���¤D�w�'¥±� ���w�'¡�¢g£���¦E��§����'¢���§�¡ý�'¡£�'������¾¦E��¦s¹�fé�&���'¡g¢����¦D¾�¥*�'�����v��¡�'¥±�����w��§�¡�'���£��?¢���'¢�'¥±� ���w�'¡�¢�£���¦E��§����'¢���§�¡ÿ�'¡£ý�'������¾¦E��¦À������§���¡�¢�¦À©±§�¥�'¯§���¢Wë�í D §�©p©±�'��¦s�á�'�D�'¢���¤D��¦z´�� D §�¤D�'¥ î Å D ¸*¹Ñr��¡�'�����D»o¢���%¡����v¯�'¥i§�©U©±�'��¦s�%¡�'�D�'¢���¤D��¦f£���%¢�§y�'¢��¢����'���w�'¡�¢y�'�v¯���� ����¢�����¦W�'���w§�¦E¢J¡�'� ����� ��¯����g��¡�¢�������¦s�f§�©�¤D�'¥±¯�'�������£�¦s¹·?¡Ð¢���À§�¢���'¥m��'¡£�»D¢���g���v¾���'¢p§�©%��¡£��'¢����'¢���£

�w§�£���© ���'¥*¦H§�©q�J¤D�'¥±¯�'�?�����£w§�¡l�'¢�¢����'���w�'¡�¢.¥*�����'�����¦á��§�¡¦E��£��'¥*�'¯�����¹��.��?�w§�¦E¢U¦E¢�¥±��Ï ��¡��w©±���'¢���¥*�&§�©Æ¢�����¦���'¥±�D�ý¦E��¯¦s�'¢Ã��¦3¢���Â��§��v¾�'¥*�'¢���¤D�¦E¾�'¥*¦s�'¡��¦s¦�§�©�w§�£���© ���'¥*¦&��¡�¢�¥*§�£������£3¯����ÌJ´7È �o¸*Î�E�ë�§���¢p§�© î Å�í´±Ä�¹ � D ¸*¹J­&¢��¨�'��§�¦s�'¥Â¦s�'¥±��¢���¡��D»f¢���º�w�,âo§�¥±��¢��ò§�©¢����¦s���Ì�É]¾���¥*��¦s��¦q�'¥*�H�'��¢���'¥?¾���¥*��¦s��§���§�� �����'�%��£�¤D�'¥*�¯����'�2�w§�£���© ���'¥*¦G´ �Ì"Ó�Ô�Í�Ô'ÇË�Ô> ±§�©����'¢���¡±»��Ô��"Ó�Ô)e�Ë�È¢  ¯���¦E��£���¦Z¡è�'¢���¹ ¸Ú§�¥£?����¦E���)�'¥±� ���w�'¡�¢�¦ �����£���£ð¯��¾�'¥±¢����'��¾����m©±§�¥±�w¦s¹%�ç�¢��'¯����D»%�����&��¡£��'¢����'¢���£J�w§�£����© ���'¥*¦J´ í î�D §�©t¢���á¢�§�¢��'��¸H�'¥*�v¦s�'�����'¢���£i¯��º¦s�'�w�'¡�¢�������'����� ����'¤ � �'¡£ ��§��v¾����'� ´ ¾§�¦s¦E��¯����£���¦s��§�¡�¢���¡��§��¦E¸J¾�¥*�'¾§�¦E��¢���§�¡¦v´ �Õ�c�Õ�ÇË�ÌUÕ¤  ��¡© ¥*§�¡�¢§�©�¡±»tÌ Ç3ÊÃÔ)¥Z¥�È3Õ¦ ±�'�v��£�¡±»�c�Ô'ÓZe�Ȧ  ¢�§�¿Ã�'¥*£�¦Z¡±»rÌ ÇË�È�ÓsÇÈÕ� ±�'¥*§���¡£�¡±»�Í�È�ÇË Ó�È� ±�'�D�'��¡¦E¢�¡±»��ÕpÛ Û Û�Õ§ ±© ¥*§��æ¹ ¹ ¹�¢�§�¡�'¢���¹ ¸*¹�­Ã¦.¢�§i¢���á¥*�'�w�'��¡���¡�����íë&��¡£��'¢����'¢���£&�w§�£����© ���'¥*¦ò´ í�B D ¸*»k¢���'�°�'¥*�ÿ��¡�¢�¥*§�£������£³¯��¨  ����� ��¢�¡¾�¥*�'¾§�¦E��¢���§�¡¦U¦E��'�z��¦?ÕW  ¢�§�¡±» Ì Çd  ��¡�¡�'¡£��ÕA ±© ¥*§��d¡±¹­&��¢��§���� �½¢�����¦Jí�B D ��§�¡�¢��'��¡¦p�m¡����v¯�'¥W§�©i£���©±© ����'����¢Ð�'¢�¢����'���w�'¡�¢�¦s»Ã§�¡�ì���'¡Ù��£��'¡�¢���© �æ¦E��¯¦s�'¢�¦§�©©±�'��¥±���v¥*�'�����'¯����?�w§�£���© ���'¥*¦D¯��w©±§��'�¦E��¡��z§�¡v¢���.¡§���¡�����£y��¡�¢�¥*§�£������£y¯���¢���i¾�¥*�'¾§�¦E��¢���§�¡»�¿&�����'�À�¦E���'�����v� ��¤D��¦G�J¦E¢�¥*§�¡��m��¡£������'¢���§�¡w§�©�¢���.¡�'¢���¥*�J§�©�¢����w§�£���© ���'¥*»Æ��¦E¾���'���'�����v��¡g¢���&����¦s�&§�©Æ�w����¦E��¥*��»�¢��'�w�¾§�¥*�'�.�'¡£%��§����'¢���¤D�f�'��¾�¥*��¦s¦E��§�¡¦s¹H�I J�I ©M�OQ�RS\ZNªVY]�\Z[S�S[S`�N�\�J�'¯����¦�ÿ¦E�§�¿Ã¦À�3¾�¥*§��v��¡�'¡�¢z��¦E� �v�w�'¢�¥±�Ò��¡Â¢���¾�¥*���'��¦E��§�¡¨§�©i��§�¡© ��¥±�w��£º�'¡£Ã¥*��¦E¢�¥±��'¢���¥*��£º�'¢�¢����'���w�'¡�¢�¦s¹«f¥*§�¡�� ¥*��¦E¢�¥±��'¢���¥*��£ �'¢�¢����'���w�'¡�¢�¦þ�'¥*��w�'��¡����ý£���m¢�§º�m�v��¦E������£���¡��k�w�'¢��'�º¯�'¢�¿Ã���'¡º¢���¾�¥*�'¾§�¦E��¢���§�¡���¡�¢�¥*§�£���'��¡��l�&µUêx�'¡£g¢���'¢?��¡�¢�¥*§�£�������¡��ì��¦E��§�¢f��¡�¢���g���'�������'�p© ¥*�'�w��§�©%��¢�¦Ã���'¡£���£��'¢�������£À´ ��B î�D ¸*¹Æ�.����¦%¢�� ¾������'�����l§����'��¥*¦%¿&��¢��M  ����� ��¢�¡¾�¥*�'¾§�¦E��¢���§�¡¦&´±��¹ �D¹D£���»D��»D�'¢���¹ ¸*¹}�ý§�¦E¢ ¡§�¢��'¯����D» ��¡½�

¥*�'���'¤D�'¡�¢w¦E��¯¦s�'¢w§�©i¢����¦s�z�v��¦E¢��'ÏD��¦s»U¢���z¤D�'¥±¯ÿ§�¥¡§���¡l�����£w¯�'��§�¡��D¦H¢�§Ð�'¡Ð��£���§��w�'¢����p�v����¢���� ¿Ã§�¥*£�'��¾�¥*��¦s¦E��§�¡¹vª)¡ ¢��� ����¦s� §�©l��§�¡© ��¥±�w��£ �'¢�¢����'���w�'¡�¢�¦s» �'¯§���¢i§�¡�&¢�����¥*£�§�©f©±�'��¦s�&¾§�¦E��¢���¤D��¦H´ � î�D ¸��¡�¤D§���¤D�¨�v����¢���� ¿Ã§�¥*£æ�'��¾�¥*��¦s¦E��§�¡¦s»y��¡æ¾�'¥±¢����'�����'¥��§��v¾§���¡£g¢��'¥±�w¦H¦E��'�l��¦�¬Ó�Ô)e�Ì �Ô'ÇË�Ô­�Ô��?Í�È�Ç�e�Ì ×É��Ì�È®  ¾�¥±���w���v��¡���¦E¢��'¥j¡±»%¿&��'¥*��¢���w¥±��� ��¢��w§�¦E¢��'������w�'¡�¢?§�©�¢���f��§��v¾§���¡£%��¦r¿&¥*§�¡�� ���3¦s�'�����'¢���£m��¦r¢��������£½§�©G¢���v���v�w��£����'¢��'���ý©±§�����§�¿&��¡��kµrµU¹rª)¡º¯§�¢��¥*��¦E¢�¥±��'¢���¥*��£À�'¡£À��§�¡© ��¥±�w��£À�'¢�¢����'���w�'¡�¢�¦s»�¢���i¥*����w�'��¡���¡�� ����¦s��¦ý´±§�¡ �'¤D�'¥*�'�D�M�2í D ¸ì�'¥*�x£���l¢�§��§��v¾����'� ¦E� ¡�¢����'¢���� ¦E¢�¥±��'¢���¥*��¦}´±��¹ �D¹l�'¾�¾§�¦E��¢���¤D���§�¡¦E¢�¥±��'¢���§�¡¦s»é��§��v¾����'�5��§�§�¥*£���¡�'¢���§�¡»x�'������¾¦E��¦�'¢���¹ ¸q¿&�����'�l�'¥*�y§���¢�¦E��£��J¢���p��§�¤D�'¥*�'�D�p§�©¢���p�'��¥*�¥*�'¡�¢�� ¥*�'�v�w�'¥*¹¯Gù"���Sö ����ø ù"�®f�'¥±�D�'¥����'������§�¡¦��'¥*�k¡§�¢Ã¡������¦s¦s�'¥±�����/¯�'¢�¢��'¥l©±§�¥¾�'¥*¦E��¡��D¹��.��#��¦s¦E��q§�© ¢���#��¡�¢��'¥±¾����'�W§�© ���'������§�¡3�'¡£� ¥*�'�v�w�'¥*»D�'��¢��§���� ��©±�'��¥±���W¿Ã�'���#��¡£��'¥*¦E¢�§�§�£W�'¢#¢������'¤D�'�.§�©�����¡�� ����¦E¢����#¢����§�¥±�D» ¦E¢������Æ¥*�'�w�'��¡¦2¢�§G¯�q© ���������¡�¤D��¦E¢����D�'¢���£v�'¢2¢���.���'¤D�'�?§�©�¾�'¥*¦E��¡��D¹�ª)¡m¢�����¦r¾�'¾�'¥*»¿Ã�y¢�¥±����£z¢�§Ð¦s�'¥*�'¢��'�Т���À¦E��¥*©±�����À§�©á¢���p¾�¥*§�¯����'�¢���¥*§���� �k�H���'¥*��© ���%�'¡�'���D¦E��¦q§�©2¢���U¾�'¥*©±§�¥±�w�'¡���i§�©�'¡l��¡�'¥*�'�w�'¡�¢��'�ã��'¾�'¡£��'¡�'�ì�'¡�'���D¦s�'¥w§�©áª)¢��'�����'¡»¿&�����'�/���'¡Ð¥±��¡Ð��¡Ð¯§�¢��/�p¡§�¡� ���'�������'����¦s��£��'¡£�����'�������'����¦s��£%�w§�£���¹�.��ò��§�¡�¢�¥±��¯���¢���§�¡ð§�©½���'�������'�l��¡©±§�¥±�w�'¢���§�¡ ¢�§

¾�'¥*¦s�3¦E�������¦s¦q��¦q��¡�'¤D�'¡����£���¦E¢�¥±��¯���¢���£À§�¤D�'¥J¯§�¢��¾�'¥±¢p§�©m¦E¾�����'�ì���'¢��'�D§�¥±����¦y�'¡£½© ¥*�'�w�v¢�� ¾��¦s¹tÑU§�¥¥*����¦s§�¡¦Ã�'¯���¡£��'¡�¢����½�������¦E¢�¥*�'¢���£3��¡ì¦s���'¢���§�¡�í»t¢���© ¥*�'�w��¦H§�©#¡§���¡w�����£�¦H�'¥*�f¡§�¢­?�����¢��p��¦U�¦s��© ���J��¦¢��§�¦s�À§�©á¤D�'¥±¯Ð�����£�¦s»%��¦E¾���'���'�����½¿&��'¡ì�'¤D�'�����'¯������¡©±§�¥±�w�'¢���§�¡l��¦3§�¡����/¦E� ¡�¢����'¢�����¹��ý§�¥*��§�¤D�'¥*»¿&���������¡©±§�¥±�w�'¢���§�¡ÿ§�¡ý¤D�'¥±¯ý¢�¥*�'¡¦E��¢���¤ ��¢��Ù§�¥º�'���'�¦s���'�w�¯��£�£���¡��H��¦U�'¥±��'���'��¢�§H© ����¢��'¥%§���¢Æ¡§���¦E�W�'¢�¢����'���w�'¡�¢�¦s»��¡©±§�¥±�w�'¢���§�¡ §�¡¢���þ�¥*�'¾§�¦E��¢���§�¡��¡�¢�¥*§�£���'��¡��l¢���§�¯�����?���z��§��v¾����'�w�'¡�¢v§�¥i¢���m��¡£���¥*���'¢v§�¯�âo���'¢v§�©i�¤D�'¥±¯À���'¡p¯�U�v��¦E������£���¡��D»��'¡£p¦E�§�����£f¢����¦¯�?�¦s��£©±§�¥ ¾�'¥*¦E��¡��v¿&��¢��v� ¥*���'¢��'¥q���'¥*��¹*�.��.�w�'��¡v¥*����¦s§�¡v��¦¢���'¢á©±�'������¥*�G¢�§J¥*�'� ��¦E¢��'¥?��¡y¢���G���'������§�¡k�'���D¾§�¦s¦E��¯����¾�¥*�'¾§�¦E��¢���§�¡¦l���'¢���'�����³©±§���¡£Ò��¡ ¥*���'�3¢��'��¢�¦z�w�'����'�¦s�g��¡£���¦E��¥*��£½§�¤D�'¥*�)© ����¢��'¥±��¡��ý§�©G�D�'¡�����¡�w�'¥±� ���w�'¡�¢�¦�´±©±�'��¦s�k¡�'�D�'¢���¤D��¦E¸*¹qª)¡x�w�'¡�� ����¦s��¦s»v�'¥±� ���w�'¡�¢m¾�¥*�'¾§�¦E��¢���§�¡¦��'¥*�����'¢���'����� ¦s�'�����'¢���£k¯��Ò¢������'�������'�#�����£W§�©t¢���m¦E��¯���'¢��'�D§�¥±��¦s��£Ã�'¥±� ���w�'¡�¢�»o¥*���¢���'¥q¢���'¡w¯��z��¢�¦i¦E��¯���'¢��'�D§�¥±��¦E��¡��3¤D�'¥±¯¹óo���v�����'¥±���D»¿&�������G��¡©±§�¥±�w�'¢���§�¡k�'¯§���¢G�'¥±� ���w�'¡�¢G§�¾�¢���§�¡�'����¢��Wc�e§�¯������D�'¢�§�¥±��¡��¦s¦&��¦Ã¦s�'��£�§��ð��§�¡© ��¥±�w��£W��¡½¥*���'�q���'¡�� ��'�D�U�¦s��»�¦E¢��'¢���¦E¢������'�t¾�¥*��©±�'¥*�'¡����¦á§�¡p¢���i§�¥*£��'¥J§�©�'¥±� ���w�'¡�¢�¥*���'����¦s�'¢���§�¡z���'¡m¯� ¤D�'¥±�H�¦s��© ����¹

��

Page 67: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

�ý§�¦E¢l�'��¥±¥*�'¡�¢w���'������§�¡¦Ò¦s�'� ¤D�'¥±� ����¢�¢����ì�'¯§���¢¢��'�v¾§�¥*�'�?�'¡£v�'��¥*�'���w¦E¢��'¡�¢����'�2�w§�£���© ���'¥*¦s»*¯���¢2�v��'��w§�¥*�º���'¡l¯�k¦s�'��£l�'¯§���¢&¢���'� ¢���'¢&��¦H�¦s��© ���&¢�§¾�'¥*¦E��¡��D¹DÑr��¥*¦E¢�»%¦s§��w�g¾�¥*�'¾§�¦E��¢���§�¡¦Ã§�¡����¨§����'��¥m¢�§��¡�¢�¥*§�£�����ý¤D�'¥±¯³�w§�£���© ���'¥*¦s¹g�.���¦s�Ù¦s�'�w�'¡�¢������'���������'¤ � ¾�¥*�'¾§�¦E��¢���§�¡¦s»�§�© ¢��'¡ ��§�¡¦E��¦E¢���¡�� §�©w�w§�¥*�¢���'¡�§�¡�?���'�������'�r��¢��'�w»�¾����'�w�&© ��¡£��'�w�'¡�¢��'�r¥*§����?��¡¢���À§�¥±�D�'¡��S°Æ�'¢���§�¡/§�©á¿&¥±��¢�¢��'¡Ð¢��'��¢�¦s»á�'¡£Ð���'¥±¢��'��¡����£���¦s�'¥±¤D�/�/¦E¾���'���'�w¾��������Ð��¡ ��¾�'¥*¦E��¡��D�)§�¥±���'¡�¢���£���'������§�¡¹­&¤D�'�����'¯�������¢��ì§�©á¢�����¦i¢�� ¾�À§�©%���'�������'�f��¡�©±§�¥±�w�'¢���§�¡ ��§�����£À¾�'¤D�â���ÿÃ�'�l¢�§�¢���º£��'¤D�'��§�¾��w�'¡�¢}§�©é¦E¾���'���'����¦s��£ Ü �v��¡���� ¾�'¥*¦s�'¥*¦sÝ §�©¢��§�¦s�¦s�'¢��'������¢��Ã�w§�£���© ���'¥*¦m¿&�§�¦s�º¦E¢�¥±��'¢���¥*�'�&¾§�¦E��¢���§�¡l��¡¢���i¦s�'¡�¢��'¡���U��¦á¦E��¯�âo���'¢t¢�§g��§�¡¦E��£��'¥*�'¯����?¤D�'¥±���'¢���§�¡¹�.���¦s�J�v��¡���� ¾�'¥*¦s�'¥*¦H��§�����£v¯�'¡��© ��¢&© ¥*§��ò��¡©±§�¥±�w���¢���§�¡Ò�'¯§���¢H¦s�'�w�'¡�¢������'�����D� ¯��¦s��£À�'����¦s¦s��¦v§�©.¡§���¡¦s»¦E��'�ÿ��¦p��§����'¢���§�¡¦s»?�w����¦E��¥*�3¢��'¥±�w¦s»&§�¥Ã¢��'�v¾§�¥*�'��'��¾�¥*��¦s¦E��§�¡¦s»#¿&�����'�x¦E�§�����£l�'��¦s§l��§�¡�¢��'��¡Ð��¡£��������¢���§�¡¨§�©U¢���m¾�¥*�'¾§�¦E��¢���§�¡º¢���'�Â�'¥*�m¢�� ¾������'�����k��¡�¢�¥*§��£������£ ¯��D¹ êD�����'¥±���D» ¢�����¦ �w§�¤D� ¥*��?�����¥*��¦�'¯�'¡£�§�¡���¡��¢���W¾�¥*�,âu�£������â���'¢i���'�������'�i��¡©±§�¥±�w���¢���§�¡ì¦E�§�����£�§�¡����¨© ��§�¿ © ¥*§�� ¢���v�����£W¢�§W��¢�¦y£����¾�'¡£��'¡�¢�¦s¹ Ñr��¡�'�����D» �'¤D�'�����'¯�������¢�� §�© ���'¥±�D�¥*�'¾�'¥±¢�§���¥*��¦�§�©J�v����¢��H¿Ã§�¥*£À��¡���¢�¦w´ ¯§�¢�� ��§��v¾����'�¾�¥*�'¾§�¦E��¢���§�¡¦��'¡£x��§��v¾§���¡£l¢��'¥±�w¦E¸�'¾�¾���'¥*¦W¢�§��'¤D�À�p���'¥±�D�p���v¾���'¢y§�¡����v¾�¥*§�¤ ��¡��½¾�'¥*¦s�g¾�¥*���'���¦E��§�¡¹�.��'¥*�q��¦?¡§3£�§���¯�¢ ¢���'¢ ��'¥±¤D��¦E¢���¡��½¦E��'�½�á¿&��£��

¥*�'¡��D�¼§�©º���'�������'����¡©±§�¥±�w�'¢���§�¡æ��¡æ¢����?���'¡�¢���¢��¡����£���£½©±§�¥W�����'��¥*�'¢��m¾�'¥*¦E��¡��k¿&�����á¥*��?�����¥*�z�'��¢��'¡�¦E��¤D�i¥*����§���¥*¦s�i¢�§p¯§�§�¢�¦E¢�¥*�'¾�¾���¡����w�'¢��§�£�¦v§�©.���'��������'��Ï ¡§�¿&����£��D�f����?�����¦E��¢���§�¡z© ¥*§��/¥*���'��¢��'��¢�¦s¹ûG�SúE�S�S�S���S�S�±�²"³µ´�¶�· ¸ ¹C¸Aº¼» ½>¾�¿"¹CÀn¸qÁ�» ½>Â�¶�¹�´�¿"â²nÄ)¹C¸ÆÅ�½>Ç�¸ ³µ³Z¿n· · ¸qÈƻɵÊ�Ë�Ë�Ê�ÌYÍ�Î�Ï>Ð�Ï"Ñ}Ò ÓnÔ�Õ�Öµ×qØÚÙ�ÛqÛqÙ�ØAÜ�Ù�Ý Ù�Õ�ÓnÏ�Ò Õ�Þ�Ô�ß�à�áÚâã Ù�ØÚáÚÒ Õ�ä¦ÔSå�æZâ Ù�Ý Ò Ù�Õ ½ª¸ ¹çÇj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é�¶�ê�´ ëC¿�ì�³ZèÆí ¹�´�¿"³Zî¹C²"´�¸ ¶�¹C²n·­ï¼¶�¹Cêð¿"³Z¿"¹CÀn¿�¶�¹ñ¾�²"¹CÄ)òC²nÄ�¿dº¼¿néÚ¶�ò�³ZÀn¿n駲"¹Cèó%ôC²n· òC²"´�¸ ¶�¹�õ}¾�²né�Ç�²n· â²néÚ½�ö÷³Z²"¹Wï¼²"¹C²"³Z¸ ²n»±ª³Z¸ éÚÀn¶�¿n½�ó�» ø)» ɵÊ�Ë�Ë�ù�Ì�ú�ØÚÔ�Ûüû�Ò Ónâ Ò Ô�Õ�Ù�Øþýÿâ Ô ÓnÔ�Ø � à�á§â ÔáÚÏnÝ åjÖµÔ�Øþä�Ù�Õ�Ò � Ò Õ�ä�û�Ò Ónâ Ò Ô�Õ�Ù�Øþý��YÝ ÏnÙ�ØÚÕ�Ò Õ�ä��nÙ�Ý ÏnÕ�Ó"ýdÙ�áÚáÚÔ�ÖÓnÒ Ù�â Ò Ô�Õ�á�Ò Õ�â Î�Ï�åjÙ�ÓnÏdÔSå��nÙ�ØÚÒ Ù�â Ò Ô�Õ Ù�Õ�û ÓnÎ�Ù�Õ�ä�Ï ½Y¸ ¹ã ØÚÔ�ÓnÏnÏnû�Ò Õ�ä�á@ÔSå�¼Ô�Ø � à�áªÐ�Ò Õ�ä�à�Ò áÚâ Ò Óná� � � � ½}¾�²"¹CÀn²néþ´�¿"³�q¹C¸ ôC¿"³ZéÚ¸ ´ �ª½�� ��» � ��î�� ��»

±ª³Z¸ éÚÀn¶�¿��÷» ½�ï¼²"³µ³Z¶�· ·�ø)» ½ ɵÊ�Ë�Ë�Ê�ÌCÞ�Ô�ß�à�áÚâ���ÓnÓnà�ØÚÙ�â Ï���â Ù�â Ò áÚÖâ Ò ÓnÙ�Ý���Õ�Õ�Ô�â Ù�â Ò Ô�Õ>ÔSåY×qÏnÕ�ÏnØÚÙ�Ý�Í�Ï"Ñ}â ½�¸ ¹ Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�鼶�ê´ ëC¿�ì�³Zè í ¹�´�¿"³µ¹C²"´�¸ ¶�¹C²n·�V�¹Cêð¿"³Z¿"¹CÀn¿�¶�¹�¾�²"¹CÄ)òC²nÄ�¿>º¼¿nîéÚ¶�ò�³ZÀn¿né�²"¹Cè�ó%ôC²n· òC²"´�¸ ¶�¹�õ}¾�²né�Ç�²n· â²néÚ½�ö÷³Z²"¹Wï¼²"¹C²"³Z¸ ²n»ï¼²"³µ³Z¶�· · ½Cø)» ½CÂ�¸ ¹�¹C¿"¹¦öq» ½C±ª³Z¸ éÚÀn¶�¿­ó�» ø)» ÉZù � � � Ì��¼Ù�ÕAáÚà�ß�ÖÓnÙ�â Ï"ä�Ô�ØÚÒ áÚÙ�â Ò Ô�Õ � ØÚÔ�ß�Ù�ß�Ò Ý Ò â Ò Ïná Î�ÏnÝ � Ù áÚâ Ù�â Ò áÚâ Ò ÓnÙ�Ý� Ù�ØÚáÚÏnØ�� ½�¸ ¹�Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é>¶�ê�´ ëC¿! �´ ë�Á�ï¼¾ "�Å�íµö$#÷Á%�

&�¶�³('CéþëC¶��ç¶�¹�ÈÆ¿"³(��¾�²"³ZÄ�¿�V�³(��¶�³Z²�õ�Â�¶�¹�´ ³Z¿n²n· ½�ï¼²"¹Cî²nè�²n» ù�ù ��î ù�Ê �»ï�ëC²"¹C¶�èçø)» Ç�» ɵÊ�Ë�Ë�ù�Ì�Þ�Ô�ß�à�áÚâ ã Ù�ØÚáÚÒ Õ�ä�Ù�Õ�ûçÜ�Ï"ý}Ô�Õ�û ½¼¸ ¹ø)» ï¼»�ø'ò�¹*)�òC²Y²"¹Cè�öq»�ôC²"¹,+ƶ�¶�³Zè É ¿nè�éÚ» ÌCÞ�Ô�ß�à�áÚâ Õ�ÏnáÚá�Ò ÕÐ�Ù�Õ�ä�à�Ù�ä�Ï Ù�Õ�û-� � ÏnÏnÓnÎ Í�ÏnÓnÎ�Õ�Ô�Ý Ô�ä�ý ½.#q¶�³Zè�³Z¿nÀ"ë�´�½/q· ò10>¿"³Z½�� ��» ù � ��î Ê�Ë 2 »3�¿nè�¿"³Z¸ Àn¸ ½WÅ�» ½WÂ�¶�¹�´�¿"â²nÄ)¹C¸ ½WÅ�» ½WÇ�¸ ³µ³Z¿n· · ¸ ½OÈÆ» ÉZù � � ��² Ì�¼Î�à�Õ 4nÒ Õ�ä�æZâ Ù�Ý Ò Ù�Õ �¦Ð�Ò Õ�ä�à�Ò áÚâ Ò Ó Ù�Õ�û Í�Ù�á54nÖµÔ�ØÚÒ ÏnÕ�â Ïnû6 �nÙ�Ý à�Ù�â Ò Ô�Õ ½�¸ ¹ Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�鼶�ê�´ ëC¿q¾�º¼ó�ï7&�¶�³('CéþëC¶��¶�¹98Zó%ôC²n· òC²"´�¸ ¶�¹ ¶�êÿÇ�²"³ZéÚ¸ ¹CÄ Å:�ªéþ´�¿"âé5;n½¤ö÷³Z²"¹C²nè�²n½Å:��²n¸ ¹C»3�¿nè�¿"³Z¸ Àn¸ ½�Å�» ½�Â�¶�¹�´�¿"â²nÄ)¹C¸ ½¼Å�» ½¼Ç�¸ ³µ³Z¿n· · ¸ ½�ÈÆ» ½¼ï¼²n· <n¶�· ²"³Z¸ ½+Æ» ÉZù � � ��= Ì$��Õ�Ù�Ý Ô�ä�ý}Öµß�Ù�áÚÏnû 6 Ñ}â ØÚÙ�Ónâ Ò Ô�Õ�ÔSå§Ð�Ï"Ñ}Ò ÓnÙ�Ý> Õ�Ô ?¼Ý Ïnû�ä�Ï@åjØÚÔ�Û@�¼Ô�Ø � Ô�ØÚÙ ��â Î�ÏA� ã ��Þ > Ð 6B6 Ñ � ÏnØÚÒ ÖÏnÕ�ÓnÏ ½�¸ ¹ Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é�¶�ê�´ ëC¿ ù éþ´�í ¹�´�¿"³µ¹C²"´�¸ ¶�¹C²n·¼ï¼¶�¹Cîêð¿"³Z¿"¹CÀn¿ ¶�¹ü¾�²"¹CÄ)òC²nÄ�¿ ³Z¿néÚ¶�ò�³ZÀn¿né�²"¹Cè®ó%ôC²n· òC²"´�¸ ¶�¹C½ö÷³Z²"¹C²nè�²n½�Å:��²n¸ ¹C»

ö÷³Z¸ éþë�â²"¹C½^º¼» ½^Â�²nÀn· ¿n¶�è~ï¼» ½ñÂ�¿C�ª¿"³Zé®Á�» ÉZù � � 2�Ì�ED�FAÐ 6HG ��ý}Õ�â Ù�Ñ���Ü�à�Ò Ý û�Ò Õ�ädÙI�¼Ô�Û � à�â Ù�â Ò Ô�Õ�Ù�Ý%Ð�Ï"Ñ}Ò ÖÓnÔ�Õ ½�¸ ¹WÇj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é�¶�êªï¼¶�· ¸ ¹CÄ ù � � 2 ½ /��ª¶�´�¶�»

ø)¿"¹CéÚ¿"¹J/q» ÉZù � � ��² Ì�æZáÚáÚà�Ïná§Ò Õ ã Ù�ØÚáÚÒ Õ�ä ½Y¸ ¹�Á�»O±�· ²néÚ¿"³É ¿nè�» Ì ½LK Ù�â à�ØÚÙ�Ý Ð�Ù�Õ�ä�à�Ù�ä�ÏYÙ�â�â Î�Ï��¼Ô�Û � à�â ÏnØ ½}Å:��³Z¸ ¹CÄ�¿"³ÈÆ¿"³Z· ²nÄ�½�±�¿"³Z· ¸ ¹C½�� ��» M�î���ì�»ø)¿"¹CéÚ¿"¹�/q» ÉZù � � ��= ÌNqÎ�ý�ÓnÔ�Û � à�â Ù�â Ò Ô�Õ�Ù�Ý}ä�ØÚÙ�ÛqÛqÙ�ØÚÒ Ù�Õ�áÓnÙ�Õ ß�Ï�á54nÏ � â Ò ÓnÙ�Ý�Ù�ß�Ô�à�â�Ï"Ñ}Ò áÚâ Ò Õ�ä¤Ý Ò Õ�ä�à�Ò áÚâ Ò Ó�â Î�ÏnÔ�ØÚÒ Ïná ½¸ ¹WÇj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é�¶�êªïEOq¾�í +Æöqî�� ��½�� ��» 2 2 ��î 2 2 ��»¾�¿"¹CÀn¸ ½qÁ�» ½Æ±�²"³µ´�¶�· ¸ ¹C¸ ½�º¼» ½�ï¼²n· <n¶�· ²"³Z¸ ½P+Æ» ½�ï¼²"³µ´�¸ ¿"³Z½�ó�»ÉµÊ�Ë�Ë�ù�Ì�QÆÔ�Ónà�ÛqÏnÕ�âR��Õ�Ù�Ý ý}áÚÒ á ½%Â�¾�íµÅ�î�M Ë�ù M�Â,�ÆÅ�íµ½ #q¿nî· ¸ ôC¿"³Z²C=�· ¿S#qì�» ù ½ »¾�¿"¹CÀn¸ ½OÁ�» ½AÂ�¶�¹�´�¿"â²nÄ)¹C¸ ½ÆÅ�» ½ÆÇ�¸ ³µ³Z¿n· · ¸ ½qÈÆ» ½ÆÅ�¶�³Z¸ ²n½Æ[ɵÊ�Ë�Ë�Ë�ÌANqÎ�ÏnØÚÏ¦Ô � � Ô�áÚÒ â Ïná¢ÛqÏnÏnâ T��U��ý}Õ�â Ù�Ónâ Ò ÓVF¦Ïnâ Ù�ÖáÚÓnÎ�ÏnÛqÏ åjÔ�ØW�¼Ô�Ø � à�áX��Õ�Õ�Ô�â Ù�â Ò Ô�Õ Ù�Õ�û ã Ù�ØÚáÚÒ Õ�ä6 �nÙ�Ý à�Ù�â Ò Ô�Õ ½�¸ ¹�Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�éO¶�ꪴ ëC¿ Ê�YLZ í ¹�´�¿"³µ¹C²"´�¸ ¶�¹C²n·ï¼¶�¹Cêð¿"³Z¿"¹CÀn¿ ¶�¹®¾�²"¹CÄ)òC²nÄ�¿¤º¼¿néÚ¶�ò�³ZÀn¿né�²"¹Cè�ó%ôC²n· òC²nî´�¸ ¶�¹C½�Á�´ ëC¿"¹CéÚ½�ö÷³Z¿n¿nÀn¿n»

Â�¶�¹�´�¿"â²nÄ)¹C¸}Å�» ½�±�²"³ZéÚ¶�´ ´�¸[3�» ½�±�²"´ ´�¸ éþ´�²�Â�» ½�ï¼²n· <n¶�· ²"³Z¸\+Æ» ½ï¼¶�³Z²R<R<n²"³Z¸]Oq» ½_^�²"Ã`��¶�· · ¸%Á�» ½�3�²"¹CÀn¸ òC· · ¸a3�» ½÷Â�²néÚéÚ¿"´�²"¹C¸Â�» ½}º¼²nêðêð²n¿n· · ¸%º¼» ½}±�²néÚ¸ · ¸%º¼» ½}Ç�²R<n¸ ¿"¹*<n²@Â�» �÷» ½}Å�²"³Z²nÀn¸ ¹C¶#q» ½]^�²"¹*<n¶�´ ´�¶V3�» ½�Â�²"¹C²b+Æ» ½�Ç�¸ ²"¹C¿néÚ¸c3�» ½a#q¿n· â¶�¹�´�¿Yº¼»ÉµÊ�Ë�Ë�Ë�ÌÆÍ�Î�Ï æZâ Ù�Ý Ò Ù�Õ���ý}Õ�â Ù�Ónâ Ò ÓnÖ(��ÏnÛqÙ�Õ�â Ò Ó�Í�ØÚÏnÏnß�Ù�Õ 4R���ØÚÓnÎ�Ò â ÏnÓnâ à�ØÚÏ õ ��Õ�Õ�Ô�â Ù�â Ò Ô�Õ õ Í�Ô�Ô�Ý á�Ù�Õ�û 6 �nÙ�Ý à�Ù�â Ò Ô�Õ ½�¸ ¹Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�éW¶�ê@´ ëC¿�ïEOq¾�í +ÆöU&�¶�³('CéþëC¶�� ¶�¹d8Z¾�¸ ¹CîÄ)òC¸ éþ´�¸ Àn²n· · � í ¹�´�¿"³(��³Z¿"´�¿nè V�³(��¶�³Z² É ¾�í +Æï¼î Ê�Ë�Ë�Ë�Ì ;n½¾�ò*e�¿"Ã`=�¶�ò�³ZÄ�½ ¼Á�òCÄ)òCéþ´ Ê�Ë�Ë�Ë ½�� ��» ù ��î Ê ��»

Çj³Z¶�À"´�¿"³Z½%Ç�» ÉZù � � � ÌjÐ�Ô�Õ�ä�ÛqÙ�Õ�QÆÒ Ónâ Ò Ô�Õ�Ù�Øþý�ÔSå��¼Ô�Õ�â ÏnÛ � Ô�ÖØÚÙ�Øþý 6 Õ�ä�Ý Ò áÚÎ ½�¾�¶�¹CÄ)â²"¹C½�¾�¶�¹Cè�¶�¹C»º�òC¸ Ã`�ª½f+Æ» ½�V�³Z²R<R<n²"³Z¸ ½bOq» ½�öq¶�· ²n½�ó�» ½�Å:��²"¹�òC½}Á�» ½�ï¼²n· î<n¶�· ²"³Z¸ ½H+Æ» ½a^�²"Ã`��¶�· · ¸ ½�Á�» ÉZù � � � Ì�Í�Î�Ï 6 à�ØÚÔ � ÏnÙ�Õ�Ð 6 ÖãHg*h1i�j�k§ã ØÚÔmljÏnÓnâ �qÍ�Î�ÏOæZâ Ù�Ý Ò Ù�Õn��ý}Õ�â Ù�Ónâ Ò ÓqÐ�Ï"Ñ}Ò ÓnÔ�Õ ½÷¸ ¹Çj³Z¶�Àn¿n¿nè�¸ ¹CÄ�é�¶�ê�´ ëC¿ ù éþ´�í ¹�´�¿"³µ¹C²"´�¸ ¶�¹C²n·�V�¹Cêð¿"³Z¿"¹CÀn¿�¶�¹¾�²"¹CÄ)òC²nÄ�¿ ³Z¿néÚ¶�ò�³ZÀn¿né ²"¹Cè ó%ôC²n· òC²"´�¸ ¶�¹C½^ö÷³Z²"¹C²nè�²n½Å:��²n¸ ¹C½ ù � � ��»

�E

Page 68: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

o�p�qsrbt5u�vxw�tsyIzf{b|s}m}s~fw�tsu�y7|s�x��:�cts�b�f~R��|1�L�(p�|E�]pm�f��rbtsy�ts�bv��Jt5u���pm�b|�Hy7|s���\�

��p�q5u��f�\��� �cox|1�\|�tsu�q5r� �b|���psqsu����\�f� �*��ts�

ox|1v:y7�:�bvf~:�d���f�f�f�:���n� �� u�p�qsr_q1tsyIzb~fqs}��:�cts�b�f~C� |1�L�(ps|sz_~:y7tsu��mpm�b|1�\�[ ¡yIp�qsu����\�f� ��¢ q1�:y�:����£,rb�:y�ts}m}�q1�:u¤u�|1��zb�:�bvf|s�_q1|�(rb�R¥f}�v�{_|t1vfvRu�|��\�\|1v

¦¨§�©�ªs«[¬a­[ª�®|Sz�u��:zb�f�L|S�mr_ts�]y7t1qsrfp��b|S��u:ts�b�(}sts��p��:�7¯ �±°�²�p��$t¥b�\|1� ¥�}Vt5zfzf}�p�q1t5�mps�:�³�¤�:u¨|s´_ts}�¥bts��pm�fµdts�bv�vf|5u�pm´�pm�fµ�mrb| vf|s´_|s}��Rzfy7|s�f� �f�·¶,� q1�:yIzb�:�b|s�f���\~|1�(zb|1q5p�t5}m}m�¨pm�±t$£,p�v�|1¸Cq1�:´_|su�tsµ_|Vts�bts}m�_�(p��`�(�_�(��|sy7¢¹ pm´_|s�º�mrb|Utsu�qsr�pm�s|1qs��¥fu�|U���d�:¥fu³��°»�(�_�(��|sy7~£,rfp�qsr7ps�Vt¼��u�ts�b�L��|5u!�(�_�(��|sy½{bt1�\|1v7�:�¾}mp��fµ�¥fp��(��p�qy7�fv:¥f}�|1�\~$q1�:u¤u�|�qs�!t5�bts}m�_�(ps�¼ps�A|s¿fzb|�qs�s|1vÀ���n{b|7tzfu�|su:|1Á:¥�p��(p���|A���Ru¨q1�Ru�u:|1qs�S��u�t5�b�(}�t5�mps�:�b~���¥fµ�µ_|1�(�mpm�fµtVq1�:u¤u�|5}�ts��p��:�A{_|s�m£n|�|s�A��rb|$��£n�f~ µ�pm´_|s�Au�|s}�t5�mpm´_|s}m�y7ts�m¥fu�|.��u�ts�b�L��|5uÂts�bvJµ_|s�b|5u�ts��p��:�Ãq1�:yIzb�:�b|s�f���\¢�®|Ä�(rb�:£º�mrfu:�:¥fµ�r�|su�u��Ru¾ts�_ts}m�_�(p��¼�mrbt5��mr_|su:|!p��pm�bvf|1|�vIt¼���mu��R�fµ¾q1�:u�u:|s}�t5�mps�:�x{b|s�m£n|1|5�I�mr_|¼Á:¥bt5}mp��m��f�,�mrb|¡��u�ts�b��}�t5��|1vÅ�:¥��mzf¥��¾ts�bv±�mrb|��(¥f{R� |1q5�mpm´_|s}m�vf|s��|5u�yIpm�b|1vxµ_�f�fv:�b|1�\�%���b�mr_|¨ts�bt5}m�_�(p��\¢¾��|¥b�\|�mrfps�%q1�:u¤u�|s}sts��p��:�.t1�%tµ�¥fp�vf|¨���:uIvf|s´_|s}��:zfy7|s�f�¼�f�t�q1�f�:u�vRpm�bt5��|1vUzbt5u�ts}�}�|5}�ts�_ts}m�_�(p��@|1�¤���Ru��Bpm�ÇÆ}�ts�fµ�¥btsµ_|1�\¢

È É�Ê�ªs«[ËaÌ�Í�­[ªsÎ(ËaÊ°Ïrb|UÁ:¥b|1���mps�:�Ð�f�±r_�:£Ñ���®��|1�����bt5�m¥fu�t5}�}sts�fµ�¥btsµ_|ts�bts}m�_�(ps�Ò�(�_�(��|sy7�Àrbt1�À{b|1|5�Óq1|s�f��u�ts}¨���Ôt5}m}A�bt5�m¥�u�ts}}�ts�fµ�¥btsµ_|À£n�:u��dpm�³�mr_|Àzbt1���%�m£n�³vf|1q�t1vf|1�L¢XÕC�Vp��7tv:p��¤� p�q5¥f}���Á:¥_|1���mp��:�_~J�¤�:uU£¼rfp�qsrÐu�|��\|1t5u�qsr_|su���rbts´_|���:¥f�bvJ�R�f}m�Åzbtsu¤�mpsts}¾ts�b�(£n|5u:�\¢½°Ïrb|¾y7�f�(�7q��:y!y7�:�ts�b�(£n|su_p��Pq1�:yIzb�:�b|s�f�c��|1���mp���µ_~:£,rb|su�|��mrb|q1�:yIzb�:�b|s�f�p��ºq1�:yIzbtsu�|1vÖtsµ_tspm�_�(��tW�(��ts�bv�tsu�vÖ���Jµ_�f�fv:�b|1�\�\~¥b�(¥bts}�}m�À�mrb|`�P|s�f�¡°�u�|�|s{bts�f�Ò���:u�×��fµ�}mp��(r�¯ �Jtsu:qs¥b�¼Ø�ÙÚfÛ ¢ ~ÝÜf���fÞ:²�~It5}m}��:£¼pm�fµ¾tÏ�f¥fy7|su�psq1ts}Ï�\q1�Ru�|�f�Ýz�u�|�qsp���p��:�ts�bv�u:|1q1t5}m}�¯�|1¢ µ_¢fw��:}�}mpm�_�\~fÜf�f��Æ:²�¢�H¥bqsr±y7|s�mrb�fvf�`rbts´_|V}�pmyIpm��t5��p��:�b�\~ar_�:£n|s´_|su�~Sts�bv

�b|1|1vÄ�s�Ä{b|���¥fzfzf}s|sy7|s�f��|1v7{f��t1vfv:pm��p��:�bt5}%y7|s�mrb�fvf�\¢� �b|Â}mpmyIpm��ts��p��:�ßps�B�mrb|Xts´_tspm}sts{fp�}mp��m�à�f�Åts�f�b�:�sts��|�vq1�:u�zb�Ru�t1~x£,rfp�q5r9vf���b�:�³|5¿fp����³���:uÔt5}m}�}�t5�fµ�¥btsµ_|1�\¢�]|1q1�:�bvR}m�_~�q1�:yIzbtsu�p��\�:�Ä���Ät5�dts�f�b�:��t5��|1vÄq1�Ru�zf¥_�xq1ts��:�f}m�xy7|1t1�(¥fu�|Ïrb�:£J£n|s}�}St¼���_�(��|sy�zfu��fv:¥bq1|1����rb|Ï��pm�bv�f��ts�_ts}m�_�(ps�%���Ru�£,r�p�qsr¡�mrb|!q��:u�zf¥b�Sp��,ts�f�b�R��t5��|1vf~_|1¢ µ_¢

}�ts{b|5}�|1vÀ{fu:t1qs�_|s�mp��fµÂ�f�x��¥fu��¤t1q1|7�(���f��ts¿b¢J×�´_ts}m¥bt5�mp��:��f�áts�bt5}m�_�(p��â�f�áv�|1|szb|5u�~ãy7�:u�|ä�\|sy7ts�f�mp�q�ts}�}m�vf|1�\q5u�pmz��mpm´_|1~I}s|s´_|s}��Òu�|�Á:¥fp�u�|��Ât1vfv:p��mp��R�bts}³ts�f�b�R��ts�s|1vq1�:u�zb�Ru�t1~ߣ,rfpsqsråy7ts�å�E�:�æ|s¿fps�(��¢ � y7�:u�|� ¥f�bvftsy7|s�f��t5}¾}mpmyIpm��t5�mps�:�-�����(¥bqsr�y7|s�mrb�fvf�Äp��Ä�mr_ts��mrb|s�Äy7|1t1�(¥fu�|¼�mrb|,µ_�f�fv:�b|1�\�V�f�t,µ�u�tsy!y7tsu¼£,pm�mrb�:¥f���ts��pm�fµApm�f���nt1q�q1�:¥f���b£,rbts�_�mr_|�µ�u�tsy!y7tsuPp���µ_�f�fvn���:u�¢°Ïr�p��n}�pmyIpm��ts��p��:��ps���:´_|su�q1�Ry7|1~�£n|dqs}�tspmy7~x�:�f}m�Â{Ý�y7|1t1�(¥fu�pm�fµÃ�mr_|Ôµ_�f�fv:�b|��\�®�f�Ât�µ�u�tsy!y7tsuÂ{f�Ãpm�s��(¥bq1q1|��\�apm�xu:|1t5}�¸ £n�:u�}�vxtszfzf}�p�q�ts�mps�:�b�\¢�®|Szfu:�:zb�f�L|S�mrbt5�]y7t1qsrfpm�b|Ï�mu:ts�b�(}sts��p��:�į �±°Ï²Ep��$t

µ_�f�fvçtszfzf}�p�q1t5�mps�:�è���ç|s´�ts}m¥bt5��|éts�bvèv:u¤pm´_|Ö�mrb|vf|s´_|s}��:zfy7|s�f�J�f�Bts�bt5}m�_�(p��Âq1�:yIzb�:�b|s�f����£,rb|5�½�mr_|�mu�ts�_�\�¤|suq1�:yIzb�:�b|s�f�]p��_{bt1�\|�v¨�:�!}mpm�fµ�¥fp����mpsqPy7�fv:¥f}�|1�\¢��¥f}��mp�¸ }mpm�fµ�¥bts}-tsz�zf}�p�q1t5�mps�:�b�W�(¥bqsrçt1�X�±°êts}�}��:£|s´_ts}m¥bts��p��:�Â�f�V�(�_�(��|syëq1�:yIzb�:�b|s�f�s�¼�mrbt5�¨�:´_|su:q1�:y7|1��mrb|Å}�pmyIpm��ts��p��:�b��y7|s�f�mp��R�b|1v�ts{b�:´_|1~Äts�bv��mrb|su�|����Ru�|�\|su�´_|1�,t1�,t`¥b�\|�� ¥f}q1�:yIzf}�|5y7|s�f�����¡�:�mrb|sux|s´_ts}m¥bt5�mps�:���|1qsrf��p�Á:¥_|1�\¢ì�,�b�R�mrb|5u-��pmµ��fp�� p�q1ts���Ót1v:´_ts�f��tsµ_|X���¥b�(pm�fµ��±°½t1�AtI�s|1���m{b|1vÒ�¤�:u%�mrb|7ts�bt5}m�_�(p��A�(�_�(�s|syWp���mrbt5�Vp��Vzfu¤p��:u¤pm��pm�c|1�7ts�_ts}m�_�(p��xzfu��:{f}s|sy7�\~Ïrfpmµ�rf}mpmµ�rf�mpm�fµ�mrb�f�\|dz�u��:{�}�|sy7���mrbts��r_ts´_|d�mrb|dµ�u�|1ts�s|1�����b|sµ_ts�mpm´_||1���¤|1qs���:�x��u�ts�_�(}sts��p��:�¾�:¥��mzf¥���¢ÕC�í�mrfp��ºzbtszb|5u�~J£n|9µ�pm´_|îts�ï�:´_|su�´�p�|s£ð�f�

¶,���]�dpm�b~ÅtÓyI¥f}m�mp�¸Ctszfzf}�p�q1t5�mp��:�Ð�bt5�m¥fu:ts}�}�t5�fµ�¥btsµ_|ts�bts}m�_�(ps��ts�bv`µ_|s�b|su�t5�mps�:�Ä�(�_�(��|sy®¥f�bvf|suvf|s´_|s}��:zfy7|s�f�ts�]�±p�qsu����\�f� �]ox|1�\|�tsu�q5rį ��|5�_�\|s�7Ø1Ù Ú�Û ¢ ~cÜf�f�fÞ�ñ ¹ tsy7�:�Ø1Ù Ú�Û ¢ ~*Üf�f�fÆ�ñ�òn|5p�vf�Ru��A�b���f�:²:~ pm�bq��:u�zb�Ru�t5�mpm�fµ�t5�bts}m�_�(ps��(�_�(��|sy7�P���:u�Æ�}�ts�fµ�¥btsµ_|1�P¯�w�rfpm�b|1�L|1~:×��fµ�}mp���rb~:ó]u�|s�_qsrb~¹ |su�y7ts�b~Ä��tszbt5�b|1�L|1~ÄôÀ�:u:|1t5�õts�bvö�Hzbts��p��(r�²�¢ � ¥�uv:p��\qs¥_�\��p��:�U����qs¥b�\|����:�Ut.v�|1�\qsu¤pmz��mp��:�U���%�mr_|¾�mr�u�|1|q1�:yIzb�:�b|s�f���V�f�E�mr_|nts�_ts}m�_�(ps�x�(�_�(��|syà¯�q�ts}�}�|1vx÷\øcØ1Ùmù5úE~û�ü:ý Ù ý[Úfþ Ù�ts�bv Ûmü:ÿ_þ ù Ú�Û��aü:ý�� ²-£,p��mr t9zbt5u���p�qs¥�}�t5u|syIzfrbt1�(p��!�:���mrb|%}��:µ�p�q1t5}x�¤�Ru�yÖvf|su�pm´_|1v±t1���mrb|À|s�bvf¸zfu��fv:¥bq5��~a£,rfpsqsrÅ�\|5u�´_|1�!t1�`��rb|Vy7|1v:pm¥fyÖ���:u,�mu�ts�b�L��|5upm�¾�:¥�ub�±°d�(�_�(��|sy7¢�®|¡ts}s�\�Aµ�pm´_|¡ts�Å�:´_|su�´�p�|s£ �f�S�mrb|¡t5u�qsr�pm�s|1qs��¥fu�|

�f�_�mr_|`�J�Hox¸ ��°®�(�_�(��|sy7~_ts�bv¡�f�_�mr_|!|s´_ts}m¥bt5�mps�:�¡£n|¥b�\|����%y7|1t1�(¥fu�|Iq1�Ru�u:|1qs���b|1�L��f���mrb|��mu:ts�b��}�t5�mp��R�b�\¢[�®|vf|sy7�:�b�(�mu�t5��|d�mrb|@q��:u¤u�|s}sts��p��R��{b|s�m£n|�|s����rb|@�Lq1�:u:|1�

���

Page 69: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

t1�\�(pmµ��b|1v%���%��u�ts�_�(}�t5��p��:���R¥f��zf¥f���¼ts�bv%�mr_|!q��:u¤u�|1q5�m�b|1�L��f�]�mrb|xts�bt5}m�_�(p��\~H¥b��pm�fµ�t1��pm}�}�¥b�(�mu:ts��p��:�n�m£n�,}�ts�fµ�¥btsµ_|1¸zbtspmu:�ts��vRp��¤��|5u�|s�����(��tsµ_|1����Pvf|s´_|s}��:zfy7|s�f�����Hzbts��p��(rb¸×��fµ�}mp��(rǯ��H×�²Â��u�ts�b��}�t5�mp��R�b~Òt1�@tÔ��|1���m{b|1v-���RuÂ�mrb|�Hzbts�fps�(r�ts�_ts}m�_�(p��¼�(�_�(��|sy7~_t5�bvVó]u�|s�_qsrb¸ ×��fµ�}mp��(r�¯ ó]×�²�mu�ts�_�(}sts��p��:�b~xt���tÄ�s|1�(��{b|1v��¤�:u¾�mr_|Äó]u�|s�_qsrÓts�_ts}m�_�(p���(�_�(��|sy7¢

��� [«��aÎ� ��XË��$ª��� I¬aÊ�¬����a©�Î�©¼­[Ë�����ËaÊ� [Ê�ªPË����� �"! Î(Ê�,�bts}m�_�(p���z�u��fv:¥_q1|1�±�mr�u�|1|Bu�|szfu:|1�\|5�f��t5�mps�:�b�Å���:u��mrb|pm�fzf¥��I�L|s�f�s|s�bq1|#���(�_|s��qsrb~azb�:u¤�mu�t5pm�xts�bvA}��:µ�p�q�ts}x���Ru�y�$�¢�H�_|s��qsrUp����mr_|.pm��pm��p�t5}7�mu:|1|�u�|szfu�|��\|s����t5�mp��:�Ã���:u��mrb|�\|s�f�s|s�bq1|�~Sts}��R�fµ�£,pm�mr±p����!t��L�\�fqspsts��|�v±ts���mu¤pm{f¥���|1¸ ´_ts}m¥b|�(�mu�¥_qs�m¥�u�|1¢H�,��|s¿_tsyIzf}�|x�f�P���_|s��qsrnp���µ�pm´_|s�npm�nó]pmµ�¥fu�|Üf~:£,rfp�qsr¡�(rb�:£n�a�mrb|`�(�_|s��qsrV��u�|1|`���:u��\|s����|s�bq�|¯¤Ü:²�¢

¯�Ü:²w�|d���:u�y7ts�Ï|��(�*z�u�ps�Ã|5�¾qsrbt5u¤µ_|�zbtsub�Bpm�bvf�:£n�]�_�f�f��mrfps�P���Ru�y7ts�*p��@��ts�_|s�Vpm�¾qsr_t5u�µ_|�{f�Â�dpm�bvf�:£n�]�_�f�f�% °�rfp��S���Ru�y7ts�*p��S�(¥fzfzb�Ru��s|1v�{��V�dpm�bvf�:£n�]�b�f���'&

(�) *,+'-/.103254'67./8�9/: ;,<';,= >�?@) ?BA,CED�0,F

�,�m��t1q5rfy7|s�f�ã�(pm�s|1�õ���RuÇz_�f�(��¸ y7�fv:p�� p�|5u��õt5u�|ë�b�:�vf|s��|5u�yIpm�b|1vãpm�á�(�_|s�sqsrb¢ ÕC�çy7�f�(�Öq1t1�\|1�\~Ô�mrb|pm�b���Ru�y7ts�mp��:�dts´_tsp�}�ts{�}�|nt1����r_|n�(���f��t1qs��p�q,��u�|1|¼p���{f¥fp�}m�p��í�b�:�ç�(¥b�¤� p�q5p�|5�f�õ��� v�|s��|5u�yIpm�b|ç£,rb|su�| |1¢ µ_¢zfu�|szb���(p��mp��R�bts}Ezfr�u�t1�L|1�,�Ru$u�|5}sts�mpm´_|Iqs}�t5¥b�\|��,�(r_�:¥f}�v%{b|ts�m�st1qsrb|�vf¢I�P�f����¸ y7�fv:p�� p�|su:�Ât5u�|³�mrf¥_�Â�(�_�(��|sy7ts�mpsq1ts}�}m�ts�m�st1qsrb|�vB���.�mr_|Åq5}��f�\|��(�Azb���\�(pm{�}�|³ts����t1q5rfy7|s�f���(p���|1~ts�bv¾u�|�ts����t1q5rb|1vf~_p����_|1q1|��\�\t5u¤�_~b{f���mrb|¨u�|1ts����t1q5rfy7|s�f�y7�fv:¥f}�|1~ft�\|5�Ï�f��rb|s¥�u�ps�(��p�q�u¤¥�}�|1�\¢ox|1ts����t1qsrfy7|s�f��u�¥f}s|1�®tsz�zf}m�ß������rb|Ó�(�_|s�sqsrÃ�s�

zfu��fv:¥bq�|�mrb|¼z_�:u¤�mu�t5pm�sñ*�mrb|z_�:u���u�tsp��¼ts�bt5}m�_�(p��V����¯�Ü:²�p��µ�pm´_|s�Jpm�Jó]pmµ�¥fu�|¾�b~P£,rb|su�|¡�mrb|¾�]�º|s¿fz�u�|1�L�(pm�fµÅ�mrb|tsµ_|s�f�Ð�f�Ó�mr_|àzbt��\��pm´_|õq��:�b�(��u�¥bq5�mp��R�b~��:u¤pmµ�pm�bt5}m}m�ts�m�st1qsrb|�vI���I�]�PÜxpm�B�(�_|s��qsr.¯¤�\|1|¼ó]pmµ�¥fu�|AÜR²`rbt��${b|1|s�u�|1t5�m��t�qsrb|1vVts�c�mrb|`�\|s����|s�_q1|E}s|s´_|s}�¢

GIH�JEKML,NOK/P@KRQTSVURSVW X�QYX,Z�S JEK[U\QEU/] ^EP@W_Pa`bX,c�dE] KeW_PafEKRNR^P@W `3L�] W_ZgW K/c,hji,dTSkPldEZ�ZgW m/W K\QTSnZgX�NoX�dTNomRdTN�NOKRQTSkc,W P@mRdEP�P@W X�QEpq X�NOK1c,KRSVU/W ] PBm/URQri�K1ZgX�dTQEc3W QrS_JEK�NOK/ZgKRNOKRQEm/K�P@p

(�) *,+'-/.1s,t�u�A,-/8�-/;,) 8E;,<';,= >�?@) ?BA,CvD�0,F

°Ïrb|�zb�Ru���u�t5pm�Ep��Ï��rb|�pm�fzf¥��E���,�mrb|Iq1�:yIzf¥f��ts��p��:�Ò�f��mrb|S}s�:µ�p�q1t5}P���Ru�yº¯ ��ó]²�~ctS}�t5{_|s}�|1v!vRpmu:|1q5��|1v`¥f�_�:u�vf|5u�|1vµ�u�tszfrßu�|sz�u�|1�L|s�f��pm�fµº�mrb|Uv�|1|szÐ�(���f��t�qs��p�qJu�|s}sts��p��:�b�tsy7�:�ݵ®�mrb|Âq1�:�f��|s���¡£n�:u�vf�.�f�A�mrb|Â�\|s����|s�bq�|¯ p�¢ |1¢ ~{bt1�(psq�zfu:|1v:psq1ts�s|1¸Ctsu�µ�¥Ýy7|s�f�Ô�(�mu�¥_qs�m¥�u�|5²�~�t5}��:�fµß£,pm�mr�\�:y7|U�\|�y7ts�f�mp�qJp��b���Ru�y7ts�mp��R�b~��(¥bqsràt1�Ô� ¥f�_qs��p��:�bt5}u�|s}sts��p��:�b�Å|s¿fz�u�|1�L�\|1vÂ{f�Wq1|5u���t5pm�Xzfu�|5zb�f��pm��p��:�b�\¢ wÄ�,��mrfps�®}�|s´_|s}�~��mr_|9v:p��¤��|5u�|s�bq�|X{b|s�m£n|1|s�ít1q5�mpm´_|9ts�bvzbt1�\��pm´_| q1�:�b�(��u�¥bq5�mps�:�b� ps�ã�b�:u�y7ts}mpm�c|1vfñëq1�:���mu:�:}u�|s}sts��p��:�b�Its�bvA}s�:�fµb¸Cv:p��(��ts�_q1|Àvf|szb|s�bv�|s�bq5p�|1�\~Ï�(¥bqsr³t1��(¥f{R�H|�qs�s�n�f�$pm�b� pm��pm��pm´_|1�\~�tsu�µ�¥fy7|s�f���nt1�\�\��qsp�t5��|1v¡£¼pm�mrµ_tszb�\~*|s��q1¢ ~ctsu:|Pu:|1�\�R}m´_|1vf¢,°Ïr_|P��ó.�f��¯�ÜR²Ïp����(rb�:£,�!pm�ó]pmµ�¥fu�|ÅÞf¢ö¶n�:��|��mr_ts�A�mr_|³�(¥fu��¤t1q1|³�(¥f{R� |1q5�����I�mrb|zbt1�\��pm´_|np��Vu�|5�bvf|5u�|1vBt1�V�mrb|yxn�:{R�Ò¯�vf|�|sz®�:{R� |�qs��²¨p����óP~fts�bv���rb| û�ÚRý�z zfr�u�t��\|t��]��rb|�xn�(¥f{¡¯�vf|1|5z¾�(¥f{R� |1q5�m²�¢

(�) *,+'-/.1{325|}(�;,<';,= >�?@) ?BA,CEDg0,F

�J�fv:ps� psq1ts��p��:�_�á��� ts�f� �f� �mrb| ts�_ts}m�_�(p��q1�:yIzb�:�b|s�f���ìt5u�| ��|��(��|�vð¥b�(pm�fµ y7�:�b�:}mpm�fµ�¥bts}u�|sµ�u�|1�\��p��:�Â� pm}s|1�Aq1�:�f�stspm��pm�fµÒ�mrb�:¥_�\ts�bv��n���Vts�bt5}m���c|1v�\|s�f�s|s�bq1|��\ñ$v:ps���¤|su�|5�bq1|��!q�ts¥b�L|1vA{f���mr_|Vy7�fv:p�� p�q1t5�mps�:�tsu�|x|s¿btsyIpm�b|1v%y7ts�f¥bts}�}m�n{f�n�mrb|�}mpm�fµ�¥fps�(�bu�|��(zb�:�b��pm{�}�|���:ub�mr_|qsrbts�fµ_|¯��H¥f��¥f��p�~:�b�f�:�f²�¢�°Ïrfps�]zfu��fq�|1�\�P�L|su�´_|1�t1�ts�npm��pm��p�t5}��\q5u�|1|s��pm�fµA���n|5�b�(¥fu�|$�mrbt5�by7�fv:p�� p�q�ts��p��:�b������mr_|ts�_ts}m�_�(ps�]rbts´_|��mr_|vf|���pmu�|1vV|1�¤��|1q5��¢

~ �����������ÕC���mrfp��I�\|1q5�mps�:��£n|,u�|s´�p�|s£Ó��rb|%{bt1��p�q��I�f���mr_|%�J�Hox¸��°ß��u�ts�_�(}sts��p��:���(�_�(��|syçts�_vBp����Ò|s´_ts}m¥bt5�mps�:�b¢ °Ïrb|u�|1t1v�|su¡p��Au:|1�¤|su¤u�|1vÒ���Ò�]p��f��rbtsy»Ø1Ù ÚfÛ�� ¯ �b�f�fÜ:²�ts�bvo�p�qsrbt5u�vf�\�:��Ø1Ù Ú�Û�� ¯ �b�f�fÜ:²I�¤�:u!� ¥�u��mr_|su!v�|s��t5pm}s�V�:�¾�mrb|ó]u�|s�bqsr�ts�bvd�Hz_ts�fps�(rd´_|su���ps�:�b�¾�f�`�mrb|±�(�_�(��|sy7¢�°Ïrb|�:´_|su�ts}m}�t5u�qsr�pm�s|1q5�m¥fu�|xts�_v,{bt��(p�qxq1�:yIzb�:�b|s�f�����mu�¥_qs�m¥�u�|

w ��� U/P�c,K/P@mRNOW i�K/coJEKRNOK�mOX�N�NOK/P_L�X�QEc,P�SVXoS JEK���� �NOKRL,NOK/P@KRQTSVURSVW X�Q5X,ZE�}U\`3L,i�K/] ]�U\Q�c3��dE�Rd,�EW����,�,�,�,��p

� �

Page 70: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

tsu�|V�mr_|¾�Ltsy7|¾���:u,{b�:��r±�mrb|Vó]×Óts�bv±�H×�´_|su���p��:�b�!�f��mrb|�(�_�(�s|sy7¢~���È���� [«��aÎ� ���J�Hox¸ ��°Ò¥b�\|1�_�mrb|P{�u��ft1v¨q��:´_|su�tsµ_|%ts�bts}m�_�(p����(�_�(��|syvf|1�\q5u�pm{_|1vÓpm�ë�]|�qs��p��:�9�b~�t®}�tsu�µ_|®yI¥f}m�mp�¸ zf¥fu�zb�f�\|�\�:¥fu�q�|1¸ }�ts�fµ�¥btsµ_|ºv:p�qs��p��:�bt5u¤�_~.t®}�|1t5u��b|�vÓ{�pm}�pm�fµ�¥bts}v:p�q5�mp��:�_tsu��_~Xts�ïtszfz�}mpsq1t5�mp��R�»pm�bv�|szb|s�_vf|s���½�stsu�µ_|s��¸}�ts�fµ�¥btsµ_|Ôµ_|s�b|su�ts��p��:�Öq1�:y!zb�:�b|s�f�Ôts�bv-t��mu:ts�b�L��|5uq1�:yIzb�:�b|s�f��¢°Ïrb|��mu�t5�b�\�¤|suÅq1�:yIzb�:�b|s����q��:�b�(ps�(���B���A�mu:ts�b�L��|su

zbts����|su��_�Uts¥����:y7ts�mp�q1t5}m}m�ît1q�Á:¥fpmu:|1v9� u:�:y �\|s�f��|s�bq1|1¸ts}mpmµ��b|1v!{fp�}mp��fµ�¥bts}%q1�:u�z_�:u�tn¯¤vf|1�\q5u�pm{_|1v!{b|s}��:£¼²¥b�(pm�fµts�ëts}�pmµ��fy7|s�f�@ts}mµ_�:u�p��mrfy vf|1�\q5u�pm{_|1vÓp��ëvf|5��t5pm}dpm��J|s�b|s�c|1��ts�_v�o�p�qsr_tsu�v��\�:�X¯ �b�f�fÜR²�¢�°�u�t5pm�fp��fµÂ��ts�_|1�zf}�t1q�|9�:�ít5}mpmµ��b|1vÐ�\|s����|s�_q1|1�®£,rfpsqsrÖrbts´_|Ó{b|�|s�ts�bts}m���c|1vÖ{f�é�mrb|ë�\�:¥fu�q1|�¸Ãts�bvÖ��tsu�µ_|s��¸ }�ts�fµ�¥btsµ_|ts�bts}m�_�(ps�ß�(�_�(��|sy7�@���-��p�|s}sv-}��:µ�p�q1t5}½���:u�y7�\¢Ò°�rb|}��:µ�p�q1t5}7���Ru�y»�(�mu�¥bq5�m¥fu:|1�\~P£¼rb|s�Uts}�pmµ��b|1vf~%t5}m}��:£ �mrb||s¿f�mu:t1q5�mp��:� �f�d}�|s¿�p�q1t5}Xts�bvö�(�mu�¥_qs�m¥�u�t5}Â��u�ts�b��}�t5�mps�:�q1�:u�u:|1�(zb�R�bvf|s�bq�|1��£,r�p�qsr»t5u�|ö�����Ru�|1vÇ���RuÓ¥b�L|öt5�u�¥f�f��pmy7|,pm�Ä�mrb|,��u�ts�b�L��|5u¨vfts�sts{bt1�\|�¢��]|1|,ó]pmµ�¥fu�| � ���Ruts�¾�:´_|su�´�p�|s£��f���mrb|E�mu:tspm��pm�fµVzfu��fq1|��\�\¢°Ïrb|Ä�mu�ts�_�\�¤|suBvft5��ts{bt��\|�p��n��u�tspm�_|1vÔ�:��Þf�f�f~ �f�f�

zbtspmu:�¼���$t5}mpmµ��b|1vn�\|s����|s�bq�|1�� u��:yWq1�:yIzf¥f��|su$y7ts�f¥bts}�����:u³�H×`~¨ts�_v®�f�f��~ �f�f�Bzbtsp�u:�.�f�Ät5}mpmµ��b|1v®w�t5�bt1vRp�ts�zbtsu�}�p�tsy7|s�f��t5u��¡vfts��t`¯ ��rb|�ònt5�b�\tsu�vVq1�:u¤zf¥b��²����Rubó]×¢

(�) *,+'-/.1�,2��n4'�"�g�n��8�-O;,) <') <'*3�':';,?@.~��l�� �a¬��(Í�¬aªsÎ(ËaʱËE� ������������]|s´_|s��|s´_ts}m¥bts���Ru��Àtsu�|�t1�(�_|�v7���B|s´_ts}m¥_ts�s|A�mr_|±�\tsy7|�\|s�d�f���\|s����|s�bq�|1�\¢xóP�RuÔ|�t1q5r9�\|s����|s�bq�|1~Vu�ts�s|su:�Jt5u�|zfu�|1�L|s�f�s|1vÄ£,pm��r®tnu:|1��|5u�|s�_q1|±�\|s�f�s|s�bq1|1~E�mrb|��Ru�pmµ�pm�_ts}×��fµ�}mp��(rU�L|s�f�s|s�bq1|�� u��:yÇ£,rfp�qsrÅ�mrb|¾rf¥fy7ts�Jó]u�|s�_qsrts�bvÄ�Hzbts�fps�(rÄ��u�ts�b��}�t5�mp��R�b��£n|su�|nvf|5u�pm´_|1vf~�ts�bv!�J�Hox¸��°¡&��¨y7t1qsrfpm�b|¾��u�ts�_�(}�t5�mps�:�b¢ ¢IÕC�X�:u�vf|suÀ����y7tspm�f��t5pm�

£ q W mRNOX,P@X,Z S¤`bU\Q,dEU/] P¥URNOK§¦ NOW S SVK\Q¨W Q¨©ªQ�«T] W_P_J¨URQEcS NOURQEP@] URSVK/c¬iª^yJEU\QEc�W QTSVX�X�S JEKRN�] URQE«�dEU�«TK/P@pB­kKjd�P@KjS_JEK/P@KS NOURQEP@] URSVW X�QEP¡U�P¡W QTLªdTS}SVX�X�dTN®Pl^EP_SVK\`�hBU\QEc¬S NOURQEP@] URSVKrS JEK\`i�U/mR�5W QTSVX3©ªQ�«T] W_P_JEp

q1�:�b�(ps�(�s|s�bqs�.tsy7�:�fµ7u�ts��|5u��P£,rb�Iy7t��¾rbts´_|Av:p�����|su:|s�f�}�|s´_|s}��`�f�S� }m¥b|s�bqs�¨pm�A�mrb|V�\�:¥�u�q1|$}�ts�fµ�¥btsµ_|1~Hu:ts�s|su��`t5u�|�b�:����rb�:£,�A��rb|V�Ru�pmµ�pm�bt5}_óau�|5�bqsr±�:u¼�Hzbts��p��(r��\|s����|s�_q1|¯����Ru���pmyIpm}�tsu.y7|s�mrb�fvf�:}s�:µ�p�|��\~7�\|�|³o$pm�fµ�µ_|suÔØ1Ù ÚfÛ ¢ ~�b�f�fÜfñ��drfp���|�Ø1Ù Ú�Û ¢ ~fÜ��f�fÞ:²�¢�,}m}@��rb|ºu:ts�s|su��Ã|s�f�s|su9�\q1�Ru�|1��u�|1� }�|�qs��pm�fµî�mrb|

ts{b�\�:}�¥f��|nÁ:¥_ts}�pm�m�B�f���mrb|¼�mu�t5�b�(}�t5�mps�:�dt1�xq1�:yIzbtsu�|1v!�s��mrb|xu�|1�¤|su�|5�bq1|x�mu:ts�b��}�t5�mp��:��µ�pm´_|s�b¢�°Ïrb|¾�:´_|su�ts}�}!�Lq1�:u:|�f�StV�\|s�f�s|s�bq1|Pp��E�mrb|Vts´_|su:tsµ_|V�f�a�mrb|V�\q1�Ru�|1�Eµ�pm´_|s�A{f��mrb|¨�L|s´_|s�¾u�t5��|su:�\¢b�]q1�Ru�|1�Su�t5�fµ_|¨� u��:y Üx��� � ~f£,pm��r.Üy7|1ts�fpm�fµY¯�° ±�²T²T³T´�µ�±�¶�·l³º¯ �_�:�Åq1�:yIzfu�|srb|s�_�(pm{�}�|5²�~I�y7|1ts�fpm�fµ�´�¸'¹O¹Oº�¶�·�»a±�²#²T³T´�µ�±/¶�·�³!¯��L�:y7|�pm�b���:u�y7ts�mp��R�xp���mu�ts�_�\�¤|su¤u�|1v¨t�q1qs¥�u�t5��|s}m��²�~*Þy7|1ts�fpm�fµn±�²#²T³T´�µ�±/¶�·�³A¯ �b�R�zb|su��¤|1q5��~¾{f¥f�@t1q�qs¥fu:ts�s|®�mu:t5�b�\��|5u��f�Ôt5}m}³pmyIzb�:u��sts�f�pm�b���Ru�y7ts�mp��:�b~Sts�bv � y7|1ts�fpm��µ�º�¼�³E±�·x¯ µ�u�tsyIy7ts�mp�q1ts}�}m�q1�:u�u:|1q5�Wts�bv ts}�}X��rb|Ãpmy!zb�:u��sts�f�Upm�_���:u�y7ts�mp��R�õps��mu�ts�_�\�¤|su¤u�|1v:²:¢

½ � ¾a¬������� [©¡��«[Ë��À¿"��¬aÊ�ÌM���ÕC���mrfp����\|1q5�mps�:��£n|ÅvRp��\q5¥b�L���(zb|�qsps� p�q³|s¿btsyIzf}�|1�À���pm}�}m¥b���mu�t5��|rb�:£Uu�|��(¥f}����%� u��:y½�±°�|s´_ts}m¥_ts��p��:�7rb|5}mz7¥_������s|1�(��ts�bvxv�|s´_|s}��:zV�mrb|`ts�bt5}��_�(p��P�(�_�(��|sy7¢½���È�¿"��ªs«[¬aÊ�©��(¬aªsÎ(ËaÊ¡Áaª��� �Â7¬aÊÏ©�¬a«[̱­[Ëa«���Í�©°Ïrb|�|s´_ts}�¥bts��p��:�d£n|�tsu:|�v:p��Lqs¥b�\��pm�fµBpm�d�mrfps�¾�\|1q5�mp��:�£nt1�`zb|su:���:u�y7|1vApm����t5�f¥bt5u����b�f�:�b~Sts�$�mrb|V{_|sµ�pm�f�fpm�fµ�f�à�:¥�u |1�¤���:u¤�Ç�:�è��rb|Ðònts�b�\tsu:vçq1�Ru�zf¥_�\¢�°Ïr_||s´_ts}m¥bts��p��:�Ó£nt1��zb|5u����Ru�y7|1vÓ�:�9t®q1�:u�zf¥b�J�f�¾�b�f��\|s�f�s|s�bq1|��\~`���`£,rfpsqsr��f�f¢ �7à ¯�Ü�Þf�B�L|s�f�s|s�bq1|��(²!£A|su�|t1�\�(pmµ��b|1vÖt9�\q��:u�|9���J�í�:uU}��:£n|5u�~JÞf�f¢ �Äà ¯�Æ ��\|s�f�s|s�bq1|��(²E£n|5u�|¼t1�L�(pmµ��b|1vIt�\q1�:u:|ϵ�u�|1ts�s|suE��rbts�I�I{f¥���b�:�`µ�u�|1ts�s|su%�mrbts�ÂÞf~$t5�bvÒÜ �Bà ¯�Þf�Ò�L|s�f�s|s�bq�|1�(²V£n|5u�|t1�\�(pmµ��b|1vVt�\q��:u�|�µ�u�|�ts�s|su_�mrbt5�¾Þf¢×�¿btsyIpm�bts�mp��R�Ö�f��ó]u:|s�bqsrÐ�L|s�f��|s�_q1|1�Bu�|1q�|spm´�pm�fµ

}��:£n¸C�\q1�:u�|¨��u�ts�b��}�t5�mp��R�b�,}�|1v¡���¾�mr_|¨psvf|s���mps� p�q�ts��p��:�Ô����\�:y7|Åqs}�t1�\�\|����f�Äts�_ts}m�_�(p��Àzfu��:{f}s|sy7�\~!�(¥bqsr½t1�¡�mr_|���:}�}��:£,pm�fµB�¸LyIp��R¸ p�vf|s�f�mps� psq1ts��p��:�¡�f�f´_�fq1t5�mp�´_|1�¸Hqs}�|�� ���a�b�:�cu�|sz�u�|��\|s����|1vxq��:u¤u�|�qs�m}m�¸LyIp��R¸Cts�bts}m�_�(p��P�f��ù1Ø¡Å7Æ þTÇ ù1Ø3Å7Æfؼ� u:|1|Eu�|5}�ts��pm´_|1�¸L{bt1v u:|szfu:|1�\|s����t5�mps�:�ì���Ñq1�:yIzf}�|s¿ pm�f´_|su:�(p��:�¯ zfu��:�_�:¥f�b¸Cvf�:¥f{f}mpm�fµ¡�f��pm�f´_|su¤��|1vx��¥f{R� |1qs��²¸L�b���mu:|1t5�my7|s�f�Ï�f��u�|1� }�|s¿�pm´_|1�¸H� p��m��|�v�zbt5u��\|��P¯ p�¢ |1¢ ~:�_�:���(zbts�f��pm�fµV�mrb|�L|s�f�s|s�bq1|5²

�J�f��� �f� �mrb| z�u��:{�}�|5y7ts�mp�q �(��u�¥bq5�m¥fu:|1� t5u�|qsrbtsu:t1qs�s|su¤p����mp�q.�f�À�(zb�:�_|s�Â}sts�fµ�¥btsµ_|Bt1�±�:zfzb�f�L|1v����y7�:u�|9���:u�y7ts}�~�£,u¤p��m��|5�í���m��}�|1�½¯ ´_�fq1t5�mpm´_|1�\~Åqs}s|1� ���\~v:pmu�|�qs��Á:¥b|1���mps�:�b��²�~]ts�bv¼rbt1v¼�b�:�b{b|�|s��|s�_q1�:¥f����|5u�|1v¼pm��:¥fu±zfu�|s´�p��:¥_�7£n�:u��_~¼£,rfp�q5r@rbt1v³pm�f´_�:}m´_|1v³y7�f�(�m}m��mu�ts�_�(}sts��p��:�®�����|�qsrf�fpsq1t5}%yÄts�f¥bts}s�\¢ � �mr_|su¨z�u��:{f}s|sy7�

�'�

Page 71: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

¯�� u:|1|Vu�|s}sts��pm´_|1�\~]u�|�� }s|s¿fpm´_|1��²ntsu�|¾t5�bts}m�_�(ps�ps�\�(¥b|���mr_ts�£n|`rbt1vV�b�:���_|s�t1v�v:u�|��\�\|�vf¢�ó]pm����|1vVzbt5u��L|1�,tsu:|`zbt5u��\|���mrbt5�Pvf���b�R�P��zbts�I�mrb|Ï£,rb�R}�|¼�\|s�f�s|s�bq1|�~�{f¥��Stsu�|Ïzfps|1q1|�v���:µ_|s�mrb|5u¨{f�B�mrb|nzbtsu��\|5u�� u:�RyÐzbtsu��mpsts}%zbt5u��\|��\ñ%� pm����|1vzbtsu��L|1�]¥b��¥bts}�}m�Vu�|1��¥f}��*pm�Vzb�f�Rub��u�ts�b��}�t5�mp��R�b�\¢×�¿btsyIzf}�|1�Ò�f�I��u�ts�b��}�t5�mps�:�b�n���:µ_|s�mrb|5u7£,pm�mr��mrb|5pmu

�\q1�:u:|Ät5u�|!µ�pm´_|s��pm�Ò°ts{f}s|!Õ:¢_°Ïr_|Ä�\�R¥fu�q�|Ä�\|5�f��|5�bq1|��tsu�|$�mrb|�óau�|s�bq5r��\|s����|s�_q1|1�\~ ��rb|�u�|���|5u�|s�bq�|x�\|s����|s�_q1|$p���mrb|¨rf¥fy7ts�.�mu�t5�b��}�ts��p��:�����¡£,rfp�qsr.�mr_|¨��u�ts�b��}�t5�mps�:�.p��q1�:yIzbtsu�|1vI{f�¾��rb|¨|s´_ts}m¥bt5���:u:�\~Ets�bvx��rb|�mu:ts�b��}�t5�mp��:�¾ps��mrb|î�R¥f�mzf¥��9���@�J�Hox¸ ��°¢Å×t1qsr �f�@�mrb|-�mr�u�|1|q1ts��|sµ_�:u¤p�|1�Òq1�R�b�(psvf|su:|1vÔts{b�R´_|�p��Àpm}m}m¥_�(��u�ts�s|1v.{f�Óts�|s¿btsyIzf}�|1¢�]|s�f��|s�_q1|¾¯ �f²À¯ £¼pm�mrJt7�Lq1�:u:|¾�f�VÜf¢ �:²%p��¨t7vRpmu�|�qs�

Á:¥b|1�(��p��:��£,p��mrÅq��:yIzf}�|s¿±pm�f´�|su���p��:�³ts�bv¨�mrb|Àvf�:¥f{�}�|1v�(¥f{R�H|�qs�$�m��zfp�q�ts}V�f�Ï�mr_ts�Vq1�:�_�(�mu�¥_qs��p��:�b¢*ÕC�±�mrb|V��óÔ���Ru¯ �f²�~ Û Øs÷ �nþ È_þ ÷HÙ ý Ø5÷IÉbØs÷ �*þ ÈbÚ/È ù1Øs÷�ps�Òts�bt5}m���c|1vÔt1�Òty7�fv:p�� p�|su�~P{b|�q1ts¥b�L|¾�mrb|7´_|su¤{ ý�Ê Æ È_þ ý ts}�u�|�t1v:�Jrbt1��t�(¥f{R�H|�qs��~1��rb|Pzfu��:�b�R¥f� þmÛ ÷ % �mr_|s�B&�¢V°Ïrb|5u�|%tsu:|%tVq1�R¥fzf}s|�f�Ät1vfvRpm��p��:�_ts}nz�u��:{�}�|sy7�¡£,p��mr���rfp��Ò�\|s����|s�_q1|T�÷ þ p��ts�bts}m���c|1v¡t1�S�mrb|!t1v:´_|su�{¡y7|1ts�fpm�fµ % �\�'&_pm�_�(�s|1t1v¡�f��t1��mrb|%q1�:�R��¥f�bq5�mps�:�¨y7|1ts�fpm�fµ % p��V&�~*ts�bv¨t%v:p�u�|1q5�$Á:¥b|1���mps�:�p��Pts�bt5}m���c|1vxt1�Ptq1��yIzf}�|sy7|��f�Ïqs}�t5¥b�\|�ñ1�mrb|�(�_|s��q5r¾ts�bv��óÇts�bt5}m�_�\|1�Â���¾�mrfps�Â�L|s�f�s|s�bq1|®tsu:|³µ�pm´_|s��pm�Ó�mr_|�,zfzb|s�bv:pm¿_¢ ¢â°Ïr_|®�J�Hox¸ ��° ��u�ts�_�(}�t5�mps�:�ë���.��rfp���\|s�f�s|s�bq1|�rbt1�®tÔ´_|su��Ã}��:£ �\q1�:u:|1~Àu:|1� }�|1q5�mpm�fµÃ�mr_|�\|s´_|su�p��m�¡�f�f�mr_|ts�bt5}m�_�(p��azfu��R{f}�|sy7�\¢°Ïrb|Ï�m£n�!�:��rb|su`�\|s����|s�bq�|1�\~c�R�!�mrb|,�R�mrb|5u�rbts�bvf~Ývf�

�b�:��rbts´_|%ts�bt5}m�_�(p��bz�u��:{�}�|sy7�O�[�mrb|Szb�f�:u��mu:ts�b��}�ts��p��:�����¯�Þ:²x¯��\q��:u�|`�b¢ Ü � ²�ps�%q�ts¥b�\|�vV{f�¡{bt1v¾ts}�pmµ��fy7|s�f�¼¯ É ý[ü�þ Ù�mu�ts�_�(}sts��|�� t1� ý[þ ÿ úbÙÂpm�b����|�t1võ�f� ÛmÚ�Ë ²�~®ts�bvà��rb|�mu�ts�_�(}sts��p��:�¾���b¯ � ²�¯��\q��:u�|ÞR²bps�]�b�:�Ïq1�:yIzf}�|5��|s}m�¡� }�¥b|s����~{f¥f���mr�p���p��!v:¥b|%���±ts�±×��fµ�}mps�(r±µ_|s�b|su�t5�mp��:�±z�u��:{�}�|sy7~u�ts�mr_|suS�mrbts�n�s�nt�ó]u�|s�_qsr�ts�bt5}m�_�(p���zfu��R{f}�|sy7¢H°Ïr�p���}st1�(��\|s�f�s|s�bq1|®ps�Â��rb|@y7�f�(��q��Ru�u�|�qs�Å£¼pm�mrëtszfz�u��:z�u�p�t5��|}�|s¿�p�q1t5}*p���|sy7�Pts�bv�rbt��]��rb|�r�p�µ�rb|1�(�Ï�\q1�Ru�|���f��rb|E�mrfu:|1|1¢� �`�mrb|±Ü�Þf�B�\|s����|s�_q1|1�V£¼pm�mrÔ�\q1�:u:|A���:uI}��:£n|su:~

ÆfÞ Ã £n|su:|7v:¥b|x���Òts�bt5}m�_�(p��zfu��:{f}s|sy7�\~$ts�bvÀ� �Äà �s�ts}mpmµ��fy7|s�f�nz�u��:{f}s|sy7�\¢`�J�f���±�f�I�mr_|�u�|1���nr_t1v.{f¥fµ_�u�|s}sts��|�v������mrb|�}s|1t5u��b|1vxv:psqs��p��R�btsu��_¢�°Ïrb|su:|�£n|su�|t��|s£q1t1�\|����f�c´_|su��Ä� u�|1|S��u�ts�b��}�t5�mps�R�b�\~1£,rb|5u�|S�mrb|Su�|1�¤|su:|s�bq1|�mu�ts�_�(}sts��p��:�À£nt��Ï´_|su�����t5u%� u��RyX�mrb|�ó]u�|s�_qsrÒ�L|s�f�s|s�bq�|1~ts�bv7�:¥�u`��u�ts�b��}�t5�mp��R�b~Ý{bt��\|1v¾�:�7�mrb|¨�\�:¥�u�q1|¨�\|s����|s�bq�|1~£nt1�]�mr_|su�|����:u:|�zb|s�_ts}�pm�c|1vf¢°Ïrb|��\|º� pmµ�¥fu�|1�U�(rb�:£ �mr_ts��~�ts�³�mrfp��U����tsµ_|º�f�

vf|s´_|s}��:zfy7|s�f�,���`�:¥�uI�(�_�(��|sy7~Ýy7�f�(�,�f�b��rb|zfu��:{�}�|sy7�pm�¡��u�ts�b��}�t5�mps�:��q1�:y7|!� u��:y9ts�bts}m�_�(p��\¢[°�u�ts�_�(}sts��p��:��q�ts�{b|ëpmyIzfu��:´_|1ví{f�ï��t1qs��}mp��fµ ts�bts}m�_�(p��àzfu��:{�}�|sy7�|s¿frfpm{�pm�s|1vd{f���mrb|�}s�:£n|1���Ò�Lq1�:u�p��fµ½�\|s�f��|5�bq1|1�L~¨ts�bvf~q1�:�f´_|su��\|5}m�_~`ts�bts}m�_�(p��%p��\��¥b|1�Àq1ts�B{b|±v:ps�\q1�:´_|su�|�v7{f�

}��f�:��pm�fµ¾ts�a��rb|¼�L|s�f�s|s�bq1|���£,p���rI�mrb|�}��:£n|��(�*�mu�t5�b�(}sts��p��:��\q1�:u:|1¢°Ïrb|Ï�_|s¿f�S�L|1q5�mp��:�Iµ�pm´_|1�$|s¿bt5yIzf}�|1�$�f�Ýp��\�(¥_|1��£,p��mr

�mrb|¨�H×��(�_�(��|sy7~f£,rfp�qsr¡p��SyÄ�:u�|y7ts�m¥fu:|��rbts�¡�mrb|`ó]×�(�_�(��|sy7¢

½��l�����ªs«[¬aÊ�©��(¬aªsÎ(ËaÊ¡Á�� [­���Ê�Î(­H¬��B��¬aÊ�Í�¬��(©�,�³|s´_ts}m¥bts��p��:�d�f���mrb|n�Hzbts��p��(rb¸ ×��fµ�}mp��(r��±°U�(�_�(��|sy£nt1�$ts}��\��zb|5u����Ru�y7|1v�pm�I�[ts�f¥_tsu��x�b�f�:�b~Ýt1� ��|5uE£n�:u��¾�:��mrb|í�±° �(�_�(��|sy rbt1vç{b|1|s�åz�u��:µ�u�|1�\��pm�fµ·���:utszfzfu��:¿�pmy7ts��|s}m�dt%�_|1tsuAts�bvÄt%rbts}s��¢Ò°�rb|À�H×U�(�_�(��|sy£nt1��vf|s´_|s}��Rzb|1vöts�bv½��|��(��|�vº¥b��pm�fµ t-q1�Ru�zf¥_���f��\|s�f�s|s�bq1|��Ä� u��:yÇ��psqsu����\�f� �x��|1qsrf��p�q�ts}xy7ts�f¥bts}��\¢½��\|s�d�f� � ���Â¥f�b�\|1|5�W�\|s����|s�_q1|1�±£nt1�±¥b�\|�vX���:u���rb||s´_ts}m¥bts��p��:�b¢� ¥f���f�Pt����R��t5}��f� � �f�A�\|s����|5�bq1|1�\~H��rb|��f¥fyI{b|su%�f�

�\|s�f�s|s�bq1|��b£,p��mrÄt%�Lq1�:u:|,� u��RyßÞ��� � £nt��b�b�fÜ!¯ � � à ²�~�mrb|�f¥fyI{b|suI�f���\|s����|s�bq�|1�P£¼pm�mr.t!�\q1�Ru�|µ�u�|1t5��|su��mrbts��Ð{f¥f��}�|1�L���mrbt5�éÞߣnt1�½Üf� � ¯�ÞfÜ Ã ²�~Åts�bvß�mr_|u�|sy7tspm�fpm�fµ³Ü � Þ��\|s����|s�_q1|1�\~ϯ �bÆ Ã ²¼rbt1v�tÀ�\q1�Ru�|À�f�Ï��:u,}��:£n|5u�¢ � �S�mr_|1�\|¡Ü � Þ��\|5�f��|s�bq�|1�`£,p��mr±�mr_|V}��:£n|��(��\q1�:u:|1�\~a�f� à ¯¤�:�±�\|s����|s�_q1|1��²Ïrbt1vAts�bt5}m�_�(p��Ez�u��:{�}�|sy7�\~ts�bv¡ÜfÆ Ã �f���mr_|sy9¯ �b�¡�L|s�f�s|5�bq1|1��²�r_t1vÀ� p��m�s|1vVz_tsu��L|1�\¢�9��|s£9�����mrb|n� pm����|1v!zbt5u��\|��\~�Æ��L|s�f�s|s�bq�|1�I�R¥f�%�f���b�f~rbt1v¾��ts¥�}m�m�¾pm��zf¥f��~b|1¢ µ_¢Ýp��fzf¥��Ï�mrbt5�¼q1�:����tsp��b|1vx¥f�f¥b��¥bts}qsrbtsu:t1qs�s|su:�9�:uºzf¥f�_qs�m¥_ts��p��R�b~d�m��zb�f�\~®�:uà�\|5�f��|5�bq1|� u�tsµ�y7|s�f���\¢°Ï��zfp�q1t5}dts�bt5}m�_�(p���z�u��:{�}�|sy7���mr_ts�Ä}�|1vÂ���Âzb�f�:u

�mu�ts�_�(}sts��p��:�Vpm�x��rb|�H×d�(�_�(�s|5yÔpm�bqs}m¥bvf|E�mrb|`���:}�}��:£,pm�fµB�¸Lpm�bq1�Ru�u�|�qs�Óts�bt5}m�_�(p��º���Ut5u�µ�¥fy7|s�f���®pm�Ðu�|s}sts��pm´_|qs}�ts¥_�\|1�\~ |1�(zb|�qsp�t5}�}m� �mrb���\|î£,p��mråt»�(pm�fµ�}�|tsu�µ�¥fy7|s�f�ϯ�ts�bvxtEzb�f�\��pm{�}�|��_�:�b¸C��´_|su��Ï�(¥f{R� |1q5�m²¸H��tsp�}m¥fu:|I���Àp�v�|s�f��p�� ���mrb|Iu:|1�¤|5u�|s�f�!���Vqs}�pm��p�q Û Øį p�¢ |1¢Æ:÷[ÙmØ�É % �_�:¥Ä& ²_pm�VpmyIzb|su�ts��pm´_|�L|s�f��|s�_q1|1�apm�x��ó¸LyIp��R¸Cts�bts}m�_�(p�� �f� �Hzbt5�fps�(r u�|�� }�|5¿fpm´_| �:u÷[ؼq1�:�_�(��u�¥bq5�mp��R�b�]p��x��ó¸Lpm�bq1�Ru�u�|�qs���(���f��t1q5�mpsqts�bt5}m�_�(p��S�f�frb�:y7�:µ�u�tszfrb�¸Lpm�bq1�Ru�u�|�qs��ts�bt5}m�_�(p��P���bq1�f�Ru�vRp��bts��p��:�¸LyIp��R¸ p�vf|s�f�mps� psq1ts��p��:�ö�f���_�:�b¸¤�:´_|su��Â�:u®q1�:���mu��R}�}�|1v�(¥f{R�H|�qs�s�¸H� p��m��|�v�zbt5u��\|��°¼ts{�}�|ÀÕ¤ÕÄq1�:�f��tsp��b�7�LtsyIzf}�|Ò�\|s�f�s|s�bq1|��7� u��:yÐ�mrb|

�H×�|s´_ts}m¥bts��p��:�b¢±óP�Ru¨|�t1qsrÄu:�:£n~c�mrb|n�L|1q1�:�bv7q1�:}�¥fyI�v:p��(z�}�ts�_�B��rb|X� zbts��p��(rà�\�:¥�u�q1|U�\|s����|s�_q1|£¼pm�mrº��rb|u�|1�¤|su�|5�bq1|õ�\|s����|5�bq1|àpm�Ñ��rb|à�b|s¿f�Ðq1�R}m¥fyI�b~Ô�mrb|�mu�ts�_�(}sts��p��:�.z�u��fv:¥_q1|1v¡{f���mr_|¨�±°º�(�_�(��|sy p��,pm���mrb|���:¥fu¤�mrUq1�:}m¥fyI�b~Vts�bv��mr_|B�Lq��:u�|.���:uÀ�mr_|¾�mu�t5�b�(}sts��p��:�t1�\�(pmµ��b|1vU{f�Ã�mrb|Ôrf¥fy7ts�Ç|s´_ts}m¥bts���Ru��dpm�-�mrb|�}�t1���q1�:}m¥fyI�b¢

� Æ

Page 72: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

Ì Í ¸'¯�ÎT²T³ Ϭ³#Ð@³TÎ#³T°�²T³ ÑjÎE±�°�¹O·�±/µVºl¸'° Í ²T¸'ÎT³¯ �f² Ò,Ó¬Ô Õ�Ö¬×�Ø Ù�Ô Ú�Ó Ô¬Û�Ú�Ü�Ö Õ�Ý�Ú�ÜRÞ¡Û�ß�à�á_â�à�ß/Ó¬Ö Õ�Ü

ã Ó ×�Ó ÜRÔ álÕ�Üjä�Õ�Üjå,Ó ×�Ú�×�Ý�Õ�Üjä�Õ�ÜjÛ�álß�æ/Ó ×�Ý�Õ�ÜjÜRÕÜRÕ�álÚ�Ó Õ�×�Ô çlÓ Ö Ü�álÙ�à�×�Ó Ü®è�Ó Õ�á¡Û�ß/à�áéÕ�ÜRÜ�Ú�êOÕ�áéä�ÕÜRØ Õ�×�Ô Õ�×�ä�álÕrÜRà�áÄà�× Û�álß�ë/álÚ ã1ã Õ3Ý�ß ã1ã à�× ìÜRß�à ã Õ�Ô Ô álÕíÚ�à ã Ó ×�Ó ÜRÔ álÕíä�Õ�Üvå,Ó ×�Ú�×�Ý�Õ�ÜRî

ïlð Ô è�Ú�ÔjñÄÕ�álÕy×�ß�ÔjÔ è�ÕyÝ�Ú�ÜRÕ�ÞñBè�êòñÄÕ�álÕóÔ è�Õ ð Ó ×�Ú�×�Ý�Õã Ó ×�Ó ÜRÔ Õ�álÜ5ß ð Ô è�Õ�Û�álß�æ/Ó ×�Ý�Õ�ÜÝ�ß�Ú�Ö Õ�ÜRÝ�Ó ×�ë5ê/Õ�ÜRÔ Õ�álä�Ú�ê�Ô ß5Ô á�êÚ�×�äaÝ�ß ã Õ�à�Û�ñBÓ Ô èaÚ¡ô�ß�Ó ×�ÔÛ�álß�ë/álÚ ã Ô ßöõ�álÓ ×�ë÷Ô ßöÔ è�Õð Ó ×�Ú�×�Ý�Õ ã Ó ×�Ó ÜRÔ Õ�álîTø

ù ß�ÔªñBÚ�Ü�Ô è�Õ}Ý�Ú�ÜRÕ�Ô è�Ú�ÔªÔ è�Õ�ê1è�Ú�æ�ÕÔ è�Õ ã Ó ×�Ó ÜRÔ Õ�álÜ ã Õ�ÔúñBè�êê/Õ�ÜRÔ Õ�álä�Ú�êûå,Ó ×�Ú�×�Ý�Õ�Üüß ð Ô è�ÕÛ�álß�æ/Ó ×�Ý�Õ�Ü3Ô álêOÓ ×�ëbÔ ß�Ú�ë/álÕ�Õ ß�×bÚÝ�ß ã1ã ß�רÛ�álß�ë/álÚ ã Ô ß¨õ�álÓ ×�ëå,Ó ×�Ú�×�Ý�Õ�Ü ð ß�á7Ô è�Õ ã Ó ×�Ó ÜRÔ Õ�á7Ô è�Ó ÜEÜRßÖ Ó ý/Õ�î

þ ø ÿ

¯�Þ:² ù ß�à�Ü�×�Õ[Û�ß/à�æ/ß�×�Ü�Û�Ú�Ü�Ú�Û/Û�à�êOÕ�áIÝ�Õ�Ô Ô Õã ß�Ô Ó ß�×�Ú�Û�á���Ü3â�à�Õ Ö Õ���Ö ß�Ý â�à�Ù�õ/Ù�Ý�ß�Ó Ü3Ú�Ó ÔálÕ ð à�ÜRÙÄä�ÕÄá�Õ�Ý�ß�×�×�Ú�� Ô álÕÄÖ ÚÄÛ�álÓ ã Ú�à�Ô Ù}ä�à ä�á_ß�Ó ÔÕ�Ô#ä�ÕíÖ Úíä�Ù ã ß�Ý�álÚ�Ô Ó ÕíÛ�ß�à�árÔ ß/à�Ü/ø

� Õ Ý�Ú�×�×�ß�ÔeÜRà�Û/Û�ß�álÔeÔ è/Ó Üã ß�Ô Ó ß�×ÀÚ ð Ô Õ�á ÜRÕ�Õ�Ó ×�ë Ô è�Õ��Ö ß�Ý��Äà�Õ�õ�Õ�Ý�ß�Ó Ü�álÕ ð à�ÜRÕMÔ ßálÕ�Ý�ß�ë/×�Ó �Õ�Ô è�Õ�álà�Ö Õ�ß ð Ö Ú�ñÚ�×�ä Ô è�Õ Û�álÓ ×�Ý�Ó Û�Ö Õ ß ðä�Õ ã ß�Ý�álÚ�Ý�ê ð ß�á�Ú�Ö ÖRø

� Õ3Ý�Ú�×�×�ß�Ô�ÜRà�Û�Û/ß�álÔ�Ô è/Ó Ü ã ß�Ô Ó ß�×Ú ð Ô Õ�á5Ô è�Õ��Ö ß�Ý�Äà�Õ�õ�Õ�Ý�ß�Ó Ü�è�Ú�ÜálÕ ð à�ÜRÕ�ä3Ô ß3álÕ�Ý�ß�ë/×�Ó �Õ1Ô è�Õ1álà�Ö Õ1ß ðÔ è�ÕíálÓ ë/è�Ô#Ú�×/ä�ä�Õ ã ß�Ý�álÚ�Ý�ê ð ß�á�Ú�Ö ÖRø

� ø þ �

¯ � ² �'× Ô Ú�×�Ôaâ/à�Õ ã Õ ã õ�álÕ[ä�Õ[Ö Ø ß�Û�Û�ß�ÜRÓ Ô Ó ß�×ß ð_ð Ó Ý�Ó Õ�Ö Ö Õ�ÞrôRÕoÝ�ß�×�Ô Ó ×/à�Õ�álÚ�Ó�ä�Ø Õ��/Õ�á�Ý�Õ�á�ä�Õ�ÜÛ�álÕ�ÜRÜRÓ ß�×�Ü�ÜRà�á�Ö Õ"ë/ß�à�æ/Õ�ál×�Õ ã Õ�×�ÔvÛ�ß�à�á�â�à�Ø Ó ÖÔ Ó Õ�×�×�ÕíÜRÕ�ÜvÛ�á_ß ã Õ�ÜRÜ\Õ�ÜvìíÝ�Õ�Ô#Ù�ë/Ú�álä'ø

� ÜíÚ ã Õ ã õ�Õ�á ß ð Ô è�Õ}ß ð_ð Ó Ý�Ó Ú�Öß�Û�Û/ß�ÜRÓ Ô Ó ß�× ï ñBÓ Ö Ö3Ý�ß�×/Ô Ó ×/à�ÕÔ ß¬Û�álÕ�ÜRÜRà�álÕjÔ è�Õjë/ß�æ/Õ�ál× ã Õ�×�ÔÔ ß ð à�Ö ð Ó Ö'Ó Ô ÜÄÛ�álß ã Ó ÜRÕ�ÜÄÓ ×3Ô è�Ó ÜálÕ�ë/Ú�álä'ø

� Ü ã Õ ã õ�Õ�á÷ß ð Ô è�Õ ß ðlð Ó Ý�Ó Ú�Öß�Û�Û/ß�ÜRÓ Ô Ó ß�×/Þ ï ñBÓ Ö Ö¡Ý�ß�×�Ô Ó ×/à�Õ�Ô ßÕ��/Õ�álÝ�Ó Ü\Õ Û�álÕ�ÜRÜRà�álÕ�Ü ß�× Ô è�Õë/ß�æ/Õ�ál× ã Õ�×�Ô ð ß�á Ó Ô3Ô ßný/Õ�Õ�ÛnÓ Ô ÜÛ�álß ã Ó ÜRÕ�ÜvÓ × Ô è�Ó ÜválÕ�ë/Ú�álä'ø

�};��'= .��@2����T;��®�'= .�?BA,CE(���8�-/;,<'?@= ;,8�) A,<

Ì Í ¸'¯�ÎT²T³ Ϭ³#Ð@³TÎ#³T°�²T³ ÑjÎE±�°�¹O·�±/µVºl¸'° Í ²T¸'ÎT³¯��:² �'ÜRÔ Õ®Û�álß�Ý�Õ�ä�Ó ã Ó Õ�×�Ô ßyÜ ��Ö ßyÕ�Ü¡Ú�Û�Ö Ó Ý�Ú�õ�Ö Õ®ÜRÓÕ�ÜRÔ �ÄÕlôRÕ�Ý�à�Ô Ú�×�ä�ß1à�×�ÚÄæ/Õ�álÜRÓ ��× ä�ÕÄÓ ä/Ó ß ã ÚÄä�Õ

� Ó ×�ä�ß�ñBÜ � � � � â�à�Õ®×/ßyÝ�ß�Ó ×�Ý�Ó ä�Ú Ý�ß�×yÕ�ÖÓ ä�Ó ß ã ÚíÕ�× Õ�Ö#â�à�Õ�ä�Õ�ÜRÕ�ÕíÕ�ÜRÝ�álÓ õ�Ó á�ø

! è�Ó Ü�Û�álß�Ý�Õ�ä�à�álÕ}Ú�Û�Û�Ö Ó Õ�Ü�ß�×�Ö êÓ ð êOß�à Ú�álÕüálà�×�×�Ó ×�ë ÚÖ Ú�×�ë/à�Ú�ë/Õ æ�Õ�álÜRÓ ß�× ß ð� Ó ×�ä�ß�ñBÜ � � � � Ô è�Ú�ÔEä�ß�Õ�ÜR×�Ø Ôã Ú�Ô Ý�èYÔ è�Õ§Ö Ú�×�ë/à�Ú�ë/Õ§êOß�àñBÚ�×�Ô#Ô ß Ô êOÛ�Õ

! è�Ó Ü¬Û�álß�Ý�Õ�ä�à�álÕ�ß�×�Ö êyÚ�Û�Û�Ö Ó Õ�Ü¬Ó ðêOß�à¤Ú�álÕ�álà�×�×�Ó ×�ë[Ú � Ó ×�ä�ß�ñBÜ� � �"� Ö Ú�×�ë/à�Ú�ë/ÕBæ/Õ�álÜRÓ ß�×1Ô è�Ú�ÔTä/ß�Õ�Ü×�ß�Ô ã Ú�Ô Ý�è Ô è�ÕÄÖ Ú�×�ë/à�Ú�ë/ÕÄÔ è�Ú�ÔªêOß�àñBÚ�×�Ô#Ô ß Ô êOÛ�ÕTø

� ø #

¯ � ² $�Õ�Û�Ó Ô ÚöÕ�ÜRÔ ÕöÛ�álß�Ý�Õ�ÜRß è�Ú�ÜRÔ Úöâ/à/Õöè�Ú�ê/ÚÕ�Ö Ó ã Ó ×�Ú�ä�ßéÔ ß�ä/ß�Ü�Ö ß�Ü�Ý�ß ã Û�ß�×�Õ�×�Ô Õ@Ü�ä�Õ"álÕ�ää�Õ�ÜRä�ÕrÖ Ú�Ü1Û�á_ß�Û�Ó Õ@ä�Ú�ä�Õ�Ü1ä�Õ%$�Õ�ä�Þ è/Ú�ë/ÚrÝ�Ö Ó ÝÕ�× � Ý�Õ�Û�Ô Ú�á1êOÞíÚ¬Ý�ß�×�Ô Ó ×/à�Ú�Ý�Ó ��×/Þ�è/Ú�ë/Ú¬Ý�Ö Ó ÝÕ�×MÒ�&¡Ý�à�Ú�×�ä/ßMÜRÕ�Ö Õ�Û�álÕ�ë/à�×�Ô Õ�ÜRÓ¡ä�Õ�ÜRÕ�ÚálÕ�Ó ×�Ó Ý�Ó Ú�á�Õ�Ö#Õ�â�à�Ó Û�ß7ø

$�Õ�Û�Õ�Ú�ÔaÔ è�Ó Ü Û�á_ß�Ý�Õ�ÜRÜ à�×�Ô Ó ÖêOß�à è�Ú�æ/Õ3ä�Õ�Ö Õ�Ô Õ�ä Ú�Ö Ö�ß ð Ô è�Õ×�Õ�Ô ñBß�álýIÝ�ß ã Û�ß�×�Õ�×�Ô Ü ð álß ãù Õ�Ô ñBß�álý Û�álß�Û�Õ�álÔ Ó Õ�ÜRÞkÝ�Ö Ó Ý�ý')( Þ�Ú�×�äöÔ è�Õ�×öÝ�Ö Ó Ý�ý+*ÄÕ�ÜñBè�Õ�×�êOß�àbÚ�á�Õ¬Û�álß ã Û�Ô Õ�äbÔ ßálÕ�ÜRÔ Ú�álÔTêOß�à�á�Ý�ß ã Û�à�Ô Õ�á�ø

$�Õ�Û�Õ�Ú�Ô¬Ô è�Ó Ü�Û�álß�Ý�Õ�ÜRÜ�à�×�Ô Ó Ö¬êOß�àè�Ú�æ/Õjä�Õ�Ö Õ�Ô Õ�ä¬Ú�Ö ÖEß ð Ô è�Õ"×�Õ�Ô ñBß�álýÝ�ß ã Û�ß�×�Õ�×�Ô Ü ð álß ã Ô è�Õ ù Õ�Ô ñBß�álýÛ�álß/Û�Õ�álÔ Ó Õ�ÜRÞ êOß�à�Ý�Ö Ó Ý�ý ')( Þ Ú�×�äêOß�à"Ý�Ö Ó Ý�ý,*ÄÕ�Ü�Ô è�Õ�×"ñBè�Õ�×"Ú�ÜRý/Ó ×�ëÔ è�Ú�Ô Ô ß álÕ�ÜRÔ Ú�álÔ Ô è�ÕbÝ�ß ã Û�à�Ô Õ�á3Ó ÜñBÚ�×�Ô Õ�ä ð ß�á�è�Ó ã ø

� ø �

¯�Æ:² �'×kÕ�ÖrÜRÓ ë/à�Ó Õ@×�Ô Õ�ÕlôRÕ ã Û�Ö ßkÜRÕ ã à�Õ�ÜRÔ álÚ�Õ�Ö×�ß ã õ�álÕbä�ÕbÖ ÚbÛ�álÕ�ÜRÕ�×�Ô Ú�Ý�Ó ��× â�à�Õ5ÜRÕbÕ�ÜRÔ �ÕlôRÕ�Ý�à�Ô Ú�×�ä�ßjÕ�×jÖ Ú�æ/Õ�×�Ô Ú�×�Ú�ä�Õ}Û�álÕ�ÜRÕ�×�Ô Ú�Ý�Ó ��×Ý�ß�× ä/Ó Ú�Û�ß�ÜRÓ Ô Ó æ/Ú�Üvà�×�ß7ø

! è�Õ ð ß�Ö Ö ß�ñBÓ ×�ë Õ��/Ú ã Û�Ö Õä�Ó ÜRÛ�Ö Ú�êOÜaÔ è�ÕM×�Ú ã ÕMß ð Ô è�ÕÛ�álÕ�ÜRÕ�×�Ô Ú�Ô Ó ß�×¡Ô è�Ú�Ô Ø Ü�Ý�à�álálÕ�×�Ô Ö êálà�×/×�Ó ×�ëòÓ ×òÜRÖ Ó ä�ÕóÜRè/ß�ññBÓ ×�ä�ß�ñ�ß�×�ÕTø

ï ×�Ô è�Õ ð ß�Ö Ö ß�ñBÓ ×�ëaÕ-�/Ú ã Û�Ö Õ�Þ Ô è�Õä�Ó ÜRÛ�Ö Ú�ê�×�Ú ã Õ¬Ô è�Ú�ÔÄÓ Ürõ�Õ�Ó ×�ë5á_à�×Ó ×oÔ è�ÕkÜRÖ Ó ä�ÕkÜRè�ß�ñ¥ñBÓ ×�ä�ß�ñ¥Ó Üä�Ó ÜRÛ�Ö Ú�êOÕ�ä ï ôRß�Ó ×7ø

þ ø .

�};��'= .��/�@2����T;��®� = ./?BA,CE40��8�-�;,<'?@= ;,8�) A,<

ÕC�Ô�mrb|³|s´_ts}m¥bts��p��:��zfu���q1|1�L�\~�rf¥fy7ts��|s´_ts}m¥bts���Ru��q1�:yIzbtsu�|1v��mrb|��±°î�mu�t5�b�(}sts�mps�:�Ã�����mr_|�u�|���|su:|s�bq�|�\|s�f�s|s�bq1|�~:pm�V�mrb|�y7ts�f�b|5u�vf|1�Lqsu�pm{_|1v$pm�¾�]|1q5��p��:� � ¢ Üf¢×�¿btsyIzf}�|¼¯��:²�~�£,p��mr7t�\q1�:u:|¼���EÞf¢ �f~�p�}m}m¥_�(��u�t5��|1�]�mrb|

��t1q5�õ�mrbt5�érf¥fy7ts�ì|s´_ts}m¥_ts���:u:�ïq1�:�b��p�vf|5u�|1v �mr_|�mu�ts�_�(}sts��p��:� % t`�Bpm�bvf�:£A�S�b�f�f�V}�ts�fµ�¥btsµ_|`´_|su��(p��:�Ä&_���{b|.t.�(}�pmµ�rf�m}m�Å£n�:u��\|¾��u�ts�b��}sts�mps�:�J�mrbt5� % t¾}sts�fµ�¥btsµ_|´_|su��(p��:�-���À�dpm�_vf�:£n���b�f���'&Ò���:u[Æ ÈbÚ21 Ø ý ÷ þmü/È ÉbØþ É þmü��nÚ É_Ø�3 þ È É ü�Ë ÷54)60606Ýñ*rb�:£n|s´_|su`�mrb|AvRp��¤��|5u�|s�bq�|p����\�Â��}�pmµ�rf�Ät1�¨�s�Ò�b�R�!{_|.q1�:�b�(psvf|su:|1vJts�Uts�bt5}m�_�(p��zfu��:{f}s|sy7¢×�¿btsyIzf}�|�¯ � ²xpm}�}m¥b���mu:ts��|��,�mrb|Ä��t5pm}m¥�u�|!���¡psvf|s���mps� �

Æ:÷[ÙmØ�É % �_�:¥Ä&ů ¥f�bv�|su������f�fvºt1�d��rb|Ó��¥f{R� |1qs���f�±�mrb|

pmyIzb|su�ts��pm´_|s²,t1�b�mrb|Pu�|���|5u�|s���$�f�a�mr_|$z�u��:�b�:yIpm�bts}$qs}�pm��p�qÛ Ø ñ*t��]y7|s�f�mp��:�b|1vxts{b�:´_|1~:�mr�p��ap��Ptq1�:y!y7�:�¾�\�:¥fu�q1|�f�{bt1vÂ�H×���u�ts�_�(}�t5��p��:�b�\¢ß°�rb|¾}�t1���Ä|s¿_tsyIzf}�|B¯�ÆR²¡ps�±t�\|s�f�s|s�bq1|Ï£,pm��rÄt¼� p��m�s|1v`zbt5u��L|¼v:¥b|Ï���`yIp��\ts�bt5}m�_�(p�������t£n�:u�v.t��%p����,rb�:y7�:µ�u�tszfra� Æ Èbü ps�Àts�_ts}m���c|1v.t1�,�mr_|� pmu:�(��zb|5u��\�:����pm�fµ�¥f}�t5u�zfu�|1�L|s������:u�y9�f�_�mrb|`´_|su�{IÆ Èbþ ý% �H�Rpm�Ä&bp��b�(�s|1t1v¾�f��t1�P��rb|�b�:¥��oÆ Èbü % �:�b|T&¤ñ*�mrb|��óJ�f��mrfps�P�\|s����|s�_q1|Ep��]µ�pm´_|s�xpm�V�mrb|E�,zfzb|s�bv:p�¿b¢½��l~87¨Î(©�­[Í�©�©�Î(ËaÊ°Ïrb|%|s¿btsyIzf}�|1��v:ps�\qs¥b�L�\|1vpm�¨�mrfps�`�\|�qs��p��:��tsu�|P�m��zfp�q�ts}��°Ïrb|,�L|s�f�s|s�bq�|1�����Ru�£,rfp�qsr!�Å�HoI¸ ��°�zfu��fvR¥bq1|��_{b|s����|5u�mu�ts�_�(}sts��p��:�b�]��|s�bv�����{b|��mrb|¼�R�b|1��£,p��mr7��|s£A|su`ts�_ts}m�_�(ps�

� �

Page 73: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

|su�u��Ru��\~f£,rfp�}�|�mrb���\|¼£,r�p�qsr�tsu�|yIp��\ts�bts}m���c|1vI�s|s�bvx���{b|�yIp��(�mu�t5�b�(}sts�s|1vf¢ÕC�¨�mrfp��b£nts�_~*|s´_ts}m¥bts�mp��R���f�c��°J�:¥f�mz�¥f�$�L|su�´_|1��t��

�:�b|�£nts�n���%z�u�p��Ru�p��mpm�c|xts�bt5}m�_�(p��Ïz�u��:{�}�|sy7�\ñÝ��rbts�bp��\~H���vf|1qsp�v�|A£,rfpsqsr�tsy7�:�fµ.�mrb|¨y7ts�Ý�Ôv:p�����|5u�|s���nt5�bts}m�_��p��zfu��:{f}s|sy7��}�|1t1v!���!�mrb|%y7�f�(�%�\|su¤p��:¥b��zfu��:{�}�|sy7�\¢ÒóP�:u|s¿btsyIzf}�|1~f�mrb|zb�f�RuIÁ:¥bts}�pm�m���f�b�mrb|��u�ts�_�(}�t5�mps�:�.�f��¯ ��²rfpmµ�rf}mpmµ�rf���!�mrb|¾�b|�|1vJ���RuÒts�ÅpmyIzfu��:´_|1vJts�bts}m�_�(p����f�q1�:yIzf}�|s¿�pm�f´_|su���p��:��pm�Ä�mrb|,ó]u�|s�bqsr�µ�u�tsyIy7tsu�~*£,rfp�qsr£,pm}m}_�_|1|1v¼���¼{_|$pm�bq��:u�zb�Ru�t5��|�v¼pm�f����mrb|V�(�_|s��qsr�ts�bv Ç �:u��ódq1�:yIzb�:�b|s�f���\¢Ä�HpmyIpm}�tsu¤}m�_~H�mrb|�zb�f�:uS�mu:ts�b��}�ts��p��:�����¯�Æ:²àpm�bvRp�q1t5��|�� ��rb|Ç�_|1|1v ���çvf|1t5}W{b|5����|suà£,p��mrrb�:y7��µ�u�tszfrb�@pm�Ö�mrb| � zbt5�fp��(rÇy7�:u�zfrb�:}s�:µ�p�q1ts}��:u�(�_|s��qsræq��:yIzb�:�b|s�f��¢�J�:u�|�µ_|s�b|su�t5}m}m�_~`�mr_|Åts�bt5}m�_�(p������!ó]×Ðt5�bv��H×

�mu�ts�_�(}sts��p��:�¨zfu��:{�}�|sy7�Erbt1�b}s|�v¼����mr_|P}mp��������f�Ïts�bt5}m�_�(p��zfu��:{f}s|sy7�»µ�p�´_|s� pm�ì�]|1qs�mps�:�b� � ¢ Ü ts�bv � ¢ �b~u�|1�(z_|1qs��pm´_|s}m�_¢±�,�bt5}m�_�(p��$zfu:�R{f}�|sy7��p�vf|5�f��p�� p�|1vIpm�7��rfp��£nts�í�mrb|s�»{b|1q1�:y7| zfu�p��:u¤pm��p�|1�Ð�¤�:uWµ�u�tsy!y7tsu Ç ��óvf|s´_|s}��:zfy7|s�f��¢

9 : ËaÊ�­��(Í�©�Î(ËaÊ�®|!rbts´_|Ä�:¥f�m}�pm�b|1v¡rb�:£ß��r_|Ä�:¥f�mzf¥��¨�f���±°ºq1ts��{b|¥b�\|1v¨t��b��|��(�m{_|1v¨�¤�:u�}mpm�fµ�¥fp�����psq%ts�bt5}m�_�(p��_pm�¨�mr_|%�\�:¥�u�q1|}�ts�fµ�¥btsµ_|1~d�(¥fzfz�}�|sy7|s�f�mpm�fµã�:�mrb|5u®y7|s�mrb�fvf�\¢·°Ïr_|y7tspm�Ât1v:´_ts�f��tsµ_|7�f�$�mrfp��Atsz�zfu��ft1q5rb~�pm�Â�:¥fuV´�p�|s£n~�p���mrbt5�Up���rb|s}�zb�X��� zfu¤p��:u¤pm��p��c|Öts�bts}m�_�(ps�Óz�u��:{�}�|sy7�\~rfpmµ�rf}mpmµ�rf�mp��fµ½�mrb�f�L|J£,rfpsqsr½rbts´_|J�mr_|Jy7�f�(�Âv:p�u�|�qs�{b|1tsu¤pm�fµ �:� �mrb|·tszfz�}mpsq�ts�mps�:�b¯���²�~ë��rb|·q1�Ru�u�|�qs�� ¥f�bq5�mp��:��pm�fµö�f�Ä£,rfpsqsr½p��Ò�mrb|Jy7tspm�½µ_�fts}Â���Ä�mrb|�(�_�(��|sy7¢

¦¨­<;�Ê�Ë��y�� [Ì�=� ��� [Ê�ªs©°Ïr�p��Szbtszb|5u�u�|sz�u�|��\|s�����S�mrb|�£n�:u��Ò�f�by7ts�f�¡zb|1�:zf}�|`pm��mrb|±¶,���Xµ�u��:¥fz�ts�A�J�Hoxñ¼£n|Åt1qs���b�:£,}�|1v:µ_|±��rb|sp�uq1�:�f��u�pm{f¥���p��:�b�\¢

�� ��� [«� [Ê�­� [©�}U\`3L,i�K/] ] h?>}pEU\QEc�@"p���dE�Rd,�EW p��,�,�,�,p � U\QE«�dEU�«TK"A QEK\dTS NOU/]NOKRL,NOK/P@KRQTSVURSVW X�QIX,ZjPl^TQ,SVU/mRSVW mbPlS N�dEmRS dTNOK/pCB QD>}p q U/] U\�EU/h>}p�� X�NO�/K/]'URQEc q p���S dTi�K/h�K/c,P@p hª��NOX,m/K/KOc,W QE«TP}X,Z,S_JEK � W NOP_S

B QTSVKRN�QEURSVW X�QEU/] ­kX�N��EP_JEX�L X�Q �'m/U/] URi�] KFE®URS dTNOU/]� U\QE«�dEU�«TK?G"QEc,KRNOP_SVU\QEc,W Q�«Tp

�}X,] ] W Q�P@h q p�H-I�I�J,pEH�JTNOK/K¬«TKRQEK\NOURSVW fEK/h�] K"K,W m/U/] W P@K/c¡`�X,c,K/] PZgX�N1P_SVURSVW P_S@W m/U/]/L�URNOP@W Q�«Tpª��NOX,m/K/K/cTW QE«,PÄX,ZTS_JEK,L�M�S J¡�"QTQ,dEU/]q K/KRSVW Q�«rX,Z#S_JEK�� � � h q U/c�NOW c,p

N U\`bX�QEh q p h¡�}p � X,�/U\QEX,hPO�p3� W QT�,JEU\` U\QEc HípQ>}KRdTS SVKRNOpH-I�I�J,pÄ��NOU/mRSVW m/U/] K"K�L�KRNOW KRQEm/K¡¦ W S Jo«�NOU\`¡`�URN5PlJEURNOW Q�« W Q`¡d�] SVW ] W Q�«�dEU/]�E � � p%B QCRvdTNOP_SVK/W QCO�p h � K/U/m/X,mR�[�}p hjK/c,P@h��NOX,m/K/KOc,W QE«TP�XªZªS JEKj­kX�N��EPlJEX�L�X�Q q UR�EW QE«SE � �b­kX�N��Eh�®� � �}X�Q�ZgKRNOK\QEm/K/h q U/c�NOW c,p

@"K/W c,X�N�QEh N p �,�,�,�,p�B QTSVK/] ] W «TK\QTSv¦1NOW S@W Q�«¡U�P@P@W_P_SVURQEm/K/p5B QT>}pU U/] K/h,@"p q X,W P�]jU\QEcV@"p"�'X�`�KRNOP@h K/c,P@p h,@"U\QEc�i�X,X���X,ZE®URS dTNOU/] � URQ�«�dEU/«TK1��NOX,m/K/P@P@W QE«Tp

O�KRQEP@KRQEhXW"p h N pX@"K/W c,X�N�QIURQEca�'pX>}W m\JEURNOc,P@X�QEh}K/c,P@pQH-I�I�L,pE®URS dTNOU/] � U\QE«�dEU�«TK1��NOX,m/K/P@P@W QE« t H�JEK1� � E � �"� L,L,NOX,U/mRJEhR�X,P_SVX�QEh�W"] d,¦ KRNOp

q URNOmRdEP@h q p hYR�p7�'URQTSVX�NOW QEW h7URQEc q p q URNOm/W QT�EW K\¦ W m/�/pZH-I�I�L,pRvdEW ] c,W Q�« U�] URNO«TK�URQ,QEX�SVURSVK/c�m/X�N�L,dEPaX,Z�©ªQE«T] W_PlJ t S JEK� KRQTQ�H�NOK/KRi�URQT�Ep,B Q ��NOX,m/K/K/c,W QE«TP5XªZ S_JEK[L?H�P_S3�"QTQ,dEU/]q K/KRSVW Q�«rX,Z#S_JEK�� � � p

q KRQEK/�/K/P@h � p¬U\QEc��'p\>�W mRJEURN/c,P@X�QEp¬�,�,�?H�p1�]R�K/P_S^A � W NOP_S� ] W_«�Q,`�KRQTSb�®] «TX�NOW S_J,` ZgX�N �"dTSVX�`bURS@W mo©�K�S NOU/mRSVW X�Q§XªZH�NOURQEP@ZgKRN q URL,L�W Q�«TP Z NOX�`_R�W ] W QE«�dEU/]��}X�N�L�X�NOU/p`B Q��NOX,m/K/KOc,W QE«TP"XªZ S JEK U URSVU A U NOW f�KRQ q H�¦¬X�NR�EPlJEX�L�hT� � ��,�,�?H�p

� W QT�,JEU\`bhaO�p h q pb�}X�NOP_SVX�Q?Acb"]_W fEKRNOh q pb��`�KRS@P�U\QEc q p� KRS SVKRQEURNOX,h¡�,�,�?H�pd>}URL�W c U/P@P�K\`3i�] ^÷X,Z5UM] URNO«TK A�P@m/U/] K� NOKRQEmRJ?Ag©�QE«,] W PlJ q H§Pl^vP_SVK\`�p\B Qö��NOX,m/K/K/c,W QE«TP�X,Z3S_JEK�,�,�?H q H���d,`¡`�W SVp

>}W mRJEURNOc,P@X�QEho�'p ho­kp R�p U X,] U�QEh�� p q K\QEK/�/K/P URQEceO�p� W QT�,JEU\`bp �,�,�?H�p � mRJEW KRf�W Q�« m/X�`¡`�KRNOm/W U/] Acf�dEU/] W S_^S NOURQEP@] URSVW X�Q ¦ W S_J K KTU\`3L�] K"A i�U/P@K/c `bKRS_JEX,c,P@p B Q��NOX,m/K/KOc,W QE«TPBXªZ#S JEK1�,�,�?H q H���d,`¡`�W SVp

>}W Q�«,«TKRNOh ©�p W"p h q p®�}X�NOP_SVX�Q?Agb"] W fEKRNOh U\QEcC>}p �}p q X,X�NOK/p�,�,�?H�phG®P@W QE« ­kX�NOc�Ag� KRN�L�] K"KTW S_^ ZgX�N � d,SVX�`bURSVW m©ªfEU/] dEU\SVW X�Q5X,Z q U/m\JEW QEK1H�NOURQEP@] URSVW X�QEp-G"QTLªdTi�] W PlJEK/c�`bP@p

��dE�\dT��W hi@"p �T�,�,�Tp � c,KRfEK/]_X�L,`bK\QTS KRQ,fEW NOX�Q,`�K\QTS ZgX�N] URNO«TK"A�P@m/U/] K�`¡dE] SVWjAg] W Q�«�dEU/]TL�URNOP@W QE«®Pl^EP_SVK\`�P@py­kX�N��EPlJEX�LX�Q N NOU\`¡`�URNb©�QE«,W QEK/KRNOW QE«aURQ�ca©ªfEU/] dEURS@W X�QEh �Xb � B E N�,�,�,�,p

­bJEW SVK/h�O�p �'p h�H�p � p�b%k �}X�QTQEK/]_] h�URQEc � p q p �}URNO] P@X�QEpPH-I�I�L,p©ªfEU/] dEU\SVW X�Q X,ZY`bU/m\JEW QEK S NOURQEP@] URS@W X�QEplB Qm@íd,`�U\Q� U\QE«�dEU�«TK HíK/mRJTQEX,] X,«�^ t ��NOX,m/KOK/c,W QE«,P5X,Z¬U ­kX�N���P_JEX�L� �5>}��� ��p,�,��n�Ag�?H��,p

� �

Page 74: W orkshop Organisers - University of Limerick · 2011. 2. 18. · Lux, Yukiko Morimoto, Mar´ıa-EugeniaNino,˜ and Fr´ed erique´ Segond. by general wellformedness principles, this

¦����� [Ê�Ì�Î�¾

(�) *,+'-/.�o3254'67./8�9/: ;,<';,= >�?@) ?BA,CED�s,F

(�) *,+'-/.�p325|}(�;,<';,= >�?@) ?BA,CEDgs,F

(�) *,+'-/.�q325|}(�;,<';,= >�?@) ?BA,CEDcq,F

Æf�