NLP 1 주차 강의Linguistic Essentials

(Ch 3)

통계적 언어처리 • 통계 ?• 전혀 모르는 언어가 쓰인 부호를 본다고 하자 .

– 예측 ?– 압축 ?– 통계 ?

• 그러면 통계적으로 언어 문제를 해결할 수 있는가 ?– 구글은 현재 통계적 방법으로 큰 성과를 거둠– 중국어 - 영어 기계번역 : 순수 통계에서 약간 벗어남– 언어 현상을 반영한 통계적 접근

• Long distance dependency • Context-free???

Competence and Performance

• Innate Learning, Categorical Statistical – CFG (Context free grammar)

• Performance

4

The Description of Language

• Grammar• set of rules which describe what is allowable in a language

• Classic Grammars (Quirk et al.)• meant for humans who know the language• definitions and rules are mainly supported by examples• no (or almost no) formal description tools; cannot be

programmed• Explicit Grammar (CFG, LFG, GPSG, HPSG, Dependency

Grammars, Link Grammars,...)• formal description• can be programmed & tested on data (texts)

5

Levels of (Formal) Description

• 6 basic levels (more or less explicitly present in most theories):– and beyond (pragmatics/logic/...)– meaning (semantics)– (surface) syntax– morphology– Phonology( 음운론 )– Phonetics( 음성학 , 발음학 )/orthography( 정서법 , 맞춤법 )

• Each level has an input and output representation– output from one level is the input to the next (upper) level– sometimes levels might be skipped (merged) or split

6

Phonetics/Orthography• Input:

– acoustic signal (phonetics) / text (orthography)• Output:

– phonetic alphabet (phonetics) / text (orthography)• Deals with:

– Phonetics:• consonant & vowel (& others) formation in the vocal tract• classification of consonants, vowels, ... in relation to

frequencies, shape & position of the tongue and various muscles in the vocal t.

• intonation – Orthography: normalization, punctuation, etc.

7

Phonology• Input:

– sequence of phones/sounds (in a phonetic alphabet); or “normalized” text (sequence of (surface) letters in one language’s alphabet) [NB nota bene (note well): phones vs. phonemes]

• Output:– sequence of phonemes (~ (lexical) letters; in an abstract alphab

et)• Deals with:

– relation between sounds and phonemes (units which might have some function on the upper level)

– e.g.: [u] ~ oo (as in book), [æ] ~ a (cat); i ~ y (flies)

8

Morphology• Input:

– sequence of phonemes (~ (lexical) letters)• Output:

– sequence of pairs (lemma, (morphological) tag)• Deals with:

– composition of phonemes into word forms and their underlying lemmas (lexical units) + morphological categories (inflection, derivation, compounding)

– e.g. quotations ~ quote/V + -ation(der.V->N) + NNS.

9

(Surface) Syntax• Input:

– sequence of pairs (lemma, (morphological) tag)• Output:

– sentence structure (tree) with annotated nodes (all lemmas, (morphosyntactic) tags, functions), of various forms

• Deals with:– the relation between lemmas & morph. categories and the senten

ce structure– uses syntactic categories such as Subject, Verb, Object,...– e.g.: I/PP1 see/VB a/DT dog/NN ~ ((I/sg)SB ((see/pres)V (a/ind dog/sg)OBJ)VP)S

10

Meaning (semantics)• Input:

– sentence structure (tree) with annotated nodes (lemmas, (morphosyntactic) tags, surface functions)

• Output:– sentence structure (tree) with annotated nodes (autosemantic -has meanin

g in isolation - lemmas, (morphosyntactic) tags, deep functions)• Deals with:

– relation between categories such as “Subject”, “Object” and (deep) categories such as “Agent”, “Effect”; adds other cat’s

– e.g. ((I)SB ((was seen)V (by Tom)OBJ)VP)S ~ (I/Sg/Pat/t (see/Perf/Pred/t) Tom/Sg/Ag/f)

11

...and Beyond• Input:

– sentence structure (tree): annotated nodes (autosemantic lemmas, (morphosyntactic) tags, deep functions)

• Output:– logical form, which can be evaluated (true/false)

• Deals with:– assignment of objects from the real world to the nodes of the sentence struct

ure– e.g.: (I/Sg/Pat/t (see/Perf/Pred/t) Tom/Sg/Ag/f) ~

see(Mark-Twain[SSN:...],Tom-Sawyer[SSN:...])[Time:bef 99/9/27/14:15][Place:39 19’40”N76 37’10”W]ş ş

12

Phonology

• (Surface Lexical) Correspondence• “symbol-based” (no complex structures)• En.: (stem-final change)

– lexical: b a b y + s (+ denotes start of ending)– surface: b a b i e s (phonetic-related: bébì0s)

• Arabic: (interfixing, inside-stem doubling) (lit. ‘read’)– lexical: kTb+uu+CVCCVC (CVCC...vowel/consonant

pattern)– surface: kuttub

13

Phonology Examples• German (umlaut) (satz ~ sentence)

– lexical: s A t z + e (A denotes “umlautable” a)– surface: s ä t z e (phonetic: zæcƏ, vs. zac)

• Turkish (vowel harmony)– lexical: e v + l A r (←houses) b a š + l A r – surface: e v l e r (heads→ b a š l a r

• Czech (e-insertion & palatalization)– lexical: m a t E K + 0 (mothers/gen.) m a t E K + ě – surface: m a t e k (mother/dat. → m a t c e

Parts of Speech and Morphology

• Parts of Speech correspond to syntactic or grammatical categories such as noun, verb, adjective, adverb, pronoun, determiner, conjunction, and preposition.

• Word categories are systematically related by morphological processes such as the formation of plural form from the singular form.

• The major types of morphological processes are inflection, derivation and compounding.

Parts of Speech• Correspond to syntactic or grammatical

categories such as noun, verb, adjectives, prepositions….

• Word categories are systematically related by morphological processes such as the formation of plural form from the singular form, past tense from present tense.

The Parts of Speech• Noun – Refer to entities like people, places, things

or idea.• Pronoun – words that take the place of nouns. • Proper noun – names.• Determiner – describes the particular action in a

noun.• Adjective – describes the properties of nouns or

pronouns.• Verb – action in a sentence.• Adverb – describes a verb, an adjective or

another adverb.• And many more

POS Labeling• Children (NOUN) eat (VERB) sweet(ADJECTIVE)

candy(NOUN)

• The(ARTICLE) children(NOUN) ate(VERB) the(ARTICLE) cake(NOUN)

• The(ARTICLE) news(NOUN) has(AUXILIARY) been(MAIN VERB) quite(ADVERB) sad(ADJECTIVE) in(PREPOSITION) fact(NOUN) .(PERIOD)

18

Morphology: Morphemes & Order

• Handles what is an isolated form in written text• Grouping of phonemes into morphemes

– sequence deliverables → deliver, able and s (3 units)– could as well be some “ID” numbers:

• e.g. deliver ~ 23987, s ~ 12, able ~ 3456 • Morpheme Combination

– certain combinations/sequencing possible, other not:• deliver+able+s, but not able+derive+s; noun+s, but not noun

+ing• typically fixed (in any given language)

19

Morphology: From Morphemes to Lemmas &

Categories • Lemma: lexical unit, “pointer” to lexicon– might as well be a number, but typically is represented as the “ba

se form”, or “dictionary headword”• possibly indexed when ambiguous/polysemous:

– state1 (verb), state2 (state-of-the-art), state3 (government)– from one or more morphemes (“root”, “stem”, “root+derivatio

n”, ...) (derivation vs. inflection)• Categories: non-lexical

– small number of possible values (< 100, often < 5-10)

20

Morphology Level: The Mapping• Formally: A+ → 2(L,C1,C2,...,Cn)

– A is the alphabet of phonemes (A+ denotes any non-empty sequence of phonemes)

– L is the set of possible lemmas, uniquely identified– Ci are morphological categories, such as:

• grammatical number, gender, case• person, tense, negation, degree of comparison, voice, aspec

t, ... • tone, politeness, ...• part of speech (not quite morphological category, but...)

– 2(L,C1,C2,...,Cn) denotes the power set of (L,C1,C2,...,Cn)

– A, L and Ci are obviously language-dependent

21

The Dictionary (or Lexicon)

• Repository of information about words:– Morphological:

• description of morphological “behavior”: inflection patterns/classes– Syntactic:

• Part of Speech• relations to other words:

– subcategorization (or “surface valency frames”)– Semantic:

• semantic features• valency frames

– ...and any other! (e.g., translation)

22

The Categories: Part of Speech: Open and Closed Categories

• Part of Speech - POS (pretty much stable set across languages)– not so much morphological (can be looked up in a dictionary), but:– morphological “behavior” is typically consistent within a POS category– Open categories: (“open” to additions)

• verb, noun, pronoun, adjective, numeral, adverb– subject to inflection (in general); subject to cross-category derivations– newly coined words always belong to open POS categories– potentially unlimited number of words

– Closed categories: • preposition, conjunction, article, interjection, clitic, particle

– not a base for derivation (possibly only by compounding)– finite and (very) small number of words

23

The Categories: Part of Speech,Open Categories: Verbs

• Verbs: – infl. categories: person, number, tense, voice, aspect, [gender, neg.], ...– syntactic/semantic: classification:

• ordinary: (to) speak, (to) write• auxiliaries: be, have, will, would, do, go (going)• modals: can, could, may, should, must, want• phrasal: begin, end, start

– morphological classification• conjugation type: regular/irregular, (Ge.: weak/strong/irregular)

– conjugation class: (Cz.: 5 classes + ~100 combinations)

24

The Categories: Part of Speech,Open Categories: Nouns

• Nouns: infl. categories: number, [gender, case, negation, ...]– semantic classification:

• human/animal/(non-living) things: driver/bird/stone• concrete/abstract: computer/thought • common/proper: table/Hopkins

– syntactic classification: countable/unc.: book, water– morphological classification:

• pluralia/singularia tantum: data (is), police (are)• declension type (“pattern” or “class”) (Cz.: 14 basic patterns, plu

s deviations: ~300 patterns, + irregular inflection)• “adverbial” nouns: afternoon, home, east (no inflection)

25

The Categories: Part of Speech,Open Categories: Pronouns

• Pronouns: infl. categories: number, gender, case, negation; person– much like nouns (syntactic usage also similar)– (pro)noun ~ “stands for” a noun– classification (mostly syntactic/semantic):

• personal: I, you, she, she, it, we, you, they• demonstrative: this, that• possessive: my, your, her, his, its, our, their; mine, yours, ours,...• reflexive: myself, yourself, herself,..., oneself• interrogative: what, which, who, whom, whose, that • indefinite (“nominal”): somebody, something, one

– morphological classification: mostly idiosyncratic pattern

26

The Categories: Part of Speech,Open Categories: Adjectives

• Adjectives: – infl. categories: degree of comp., [number, gender, case, negation]– classification:

• ordinary: new, interesting, [test (equipment)]• possessive: John’s, driver’s• proper: Appalachian (Mountains)• often derived from verbs/nouns: teaching (assistant), trendy, stylish

– morphological classification• mostly regular declension (Cz.: 4 basic patterns, ~ 10 total)• degrees of comparison (En.: big, bigger, biggest)• but: large number of forms (agreement, cf. section on syntax)

27

The Categories: Part of Speech,Open Categories: Adverbs

• Adverbs: “infl.” categories: degree of comp., [negation]– open cat.: regular derivation from adjectives common:

• new → newly, interesting → interestingly– non-derived adverbs:

• ordinary: so, well, just, too, then, often, there• wh-adverbs (interrogative): why, when, where, how• degree adverbs/qualifiers: very, too

– morphological classification (not much, really...)• degree of comparison: well, better, best

– soon, sooner (other lang.: all 3 degrees regular)

28

The Categories: Part of Speech,Open Categories: Numerals

• Numerals: infl. categories: number, gender, case, negation– open cat.: compounding (Ge.: einundzwanzig, 21)– classification:

• cardinals: one, five, hundred– NB: million etc. often considered noun

• ordinals/fractionals: first, second, thirtieth• quantifiers: all, many, some, none• multiplicative: times, twice (Cz.: dvaadvacetkrát, 22-times)• multilateral: single, triple, twofold

– morphological classification: as nouns/adjectives; many irreg.

29

The Categories: Part of Speech, Closed Categories

• Closed categories: preposition, conjunction, article, interjection, clitic, particle– Morphological behavior: indeclinable (no declension, no conjugation)

• preposition: of, without, by, to; • conjunction: coordinating: and, but, or, however subordinating: that, if, because, before, after, although, as • article: a, the; • interjection: wow, eh, hello;• clitic: ‘s; may be attached to whole phrases (at the end)• particle: yes, no, not; to (+verb);

– many (otherwise) prepositions if part of phrasal verbs, e.g. (look) up

30

The Categories: Number and Gender

• Grammatical Number: Singular, Plural– nouns, pronouns, verbs, adjectives, numerals

• computer / computers; (he) goes / (they) go– In some languages (Czech): Dual (nouns, pronouns, adjectives)

• (Pl.) nohami / (Dl.) nohama (Cz.; (by) legs (of sth)/(by) legs (of sb))• Grammatical Gender: Masculine, Feminine, Neuter

– nouns, pronouns, verbs, adjectives, numerals• he/she/it; читал, читала, читало (Ru.; (he/she/it) was-reading)• nouns: (mostly) do not change gender for a single lexical unit

– Also: animate/inanimate (gram., some genders), etc.• Mädchen (Ge.; girl, neuter); děti (Cz.; children, masc. inanim.)

31

The Categories: Case• Case

– English: only personal pronouns/possessives, 2 forms– other languages: 4 (German), 6 (Russian), 7 (Czech,Slovak,...)

• nouns, pronouns, adjectives, numerals– most common cases (forms in singular/plural)

• nominative I/we (work) tøída/tøídy (Cz.; class)• genitive (picture of) me/us tøídy/tøíd• dative (give to) me/us tøídě/tøídám• accusative (see) me/us tøídu/tøídy• vocative -/- tøído/tøídy• locative (about) me/us tøídě/tøídách• instrumental (by) me/us tøídou/tøídami

32

The Categories: Person, Tense

• Person– verbs, personal pronouns

• 1st, 2nd, 3rd: (I) go, (you) go, (he) goes; (we) go, (you) go, (they) go• jdu, jde , jde, jdeme, jdete, jdou (Cš

z.)• Tense (Cz.: go) (Pol.: go)

– past: (you) went - szliœcie– present: (you pl.) go jdete idziecie– future (!if not “analytical”) - pùjdete -– concurrent (gerund) going jda id cą– preceding - - sze³szy

33

The Categories: Person, Tense• Person

– verbs, personal pronouns• 1st, 2nd, 3rd: (I) go, (you) go, (he) goes; (we) go, (you) go, (they) go• jdu, jde , jde, jdeme, jdete, jdou (Cš

z.)• Tense (Cz.: go) (Pol.: go)

– past: (you) went - szliœcie– present: (you pl.) go jdete idziecie– future (!if not “analytical”) - pùjdete -– concurrent (gerund) going jda id¹c– preceding - - szed³szy

34

Note on Tense• Grammars: more (syntactic/sematnic) tenses Time X

– but: morphology handles isolated words → some tenses can be defined & handled only at an upper level (surface syntax)

• Examples of (traditional) tense (synthetical and analytical):• infinitive: (to) write (tenseless, personless, ..., except negation (Cz.))• simple present/past: (I) write/(she) writes; (I,she) wrote• progressive present/past: (I) am writing; (I) was writing• perfect present/past: (I) have written; (I) had written• all in passive voice (cf. later), too:

– (the book) is being/has been/had been written etc.• all in conditional mood, too (mood: in Eng. not a morph. category!)

– (the book) would have been written

35

The Categories: Voice & Aspect

• Voice– active vs. passive

• (I) drive / (I am being) driven• (Ich) setzte (mich) / (Ich bin) gesetzt (Ge.: to sit down)

• Aspect– imperfective vs. perfective:

• пoкупал / купил (Ru.: I used to buy, I was buying) / I (have) bought)– imperfective continuous vs. iterative (repeating)

• spal / spával (Cz.: I was sleeping / I used to sleep (every ...))

36

The Categories: Negation, Degree of Comparison• Negation:

– even in English: impossible (~ not possible)• Cz: every verb, adjective, adverb, some nouns; prefix ne-

• Degree of Comparison (non-analytical):– adjectives, adverbs:

• positive (big), comparative (bigger), superlative (biggest)• Pol.: (new) nowy, nowszy, najnowszy

• Combination (by prefixing):– order? both possible: (neg.: Cz./Pol.: ne-/nie-, sup.: nej-/naj-)

• Cz.: nejnemoٱnìj í (the most impossible)š• Pol.: nienajwierniejszy (the most unfaithful)

37

Typology of Languages• By morphological features

– Analytical: using (function) words to express categories• English, also French, Italian, ..., Japanese, Chinese

– I would have been going ~ (Pol.) szłabym– Inflective: using prefix/suffix/infix, combines several categ.

• Slavic: Czech, Russian, Polish,... (not Bulgarian); also French, German; Arabic

– (Cz. new(acc.)) novou (Adj, Fem., Sg., Acc., Non-neg., Pos.)– Agglutinative: one category per (non-lexical) morpheme

• Finnish, Turkish, Hungarian– (Fin. plural): -i-

38

Categories & Tags• Tagset:

– list of all possible combinations of category values for a given language– T C1ⅹC2ⅹ... ⅹCn

– typically string of letters & digits:• compact system: short idiosyncratic abbreviations:

– NNS (gen. noun, plural)• positional system: each position i corresponds to Ci:

– AAMP3----2A---- (gen. Adj., Masc., Pl., 3rd case (dative), comparative (2nd degree of comparison), Affirmative (no negation))

– tense, person, variant, etc.: N/A (marked by “empty position”, or ‘-’)

• Famous tagsets: Brown, Penn, Multext[-East], ...

Words’ Syntactic Functions• Typically, nouns refer to entities in the world like pe

ople, animals and things.• Determiners describe the particular reference of a no

un and adjectives describe the properties of nouns.• Verbs are used to describe actions, activities and state

s.• Adverbs modify a verb in the same way as adjectives

modify nouns. Prepositions are typically small words that express spatial or time relationships. Prepositions can also be used as particles to create phrasal verbs. Conjunctions and complementizers link two words, phrases or clauses.

Syntax or Phrase Structure: A simplecontext-free grammar

• S --> NP VP• NP --> AT NNS | AT

NN | NP PP• VP --> VP PP | VBD |

VBD NP• P --> IN NP

• AT --> the• NNS --> children |

students | mountains • VBD --> slept | ate | saw• IN --> in | of• NN --> cake

The Grammar The Lexicon

Syntax or Phrase Structure: A Parse Tree

A Simple Context-Free Grammar

• The Grammar rules• S -> NP V• NP -> N

• The Lexicon• N -> John, Gaurav, Ram ……• V -> walks, talks, eats, went …..

Tag Sets• A tag indicates the various

conventional parts of speech.• Different Tag Sets have been used:

E.g., Brown Tag Set, Penn Treebank Tag Set.

• Tag examples: NP Proper noun, NN Singular noun, AT Article, DET Determinant.

Stochastic Grammars• Grammars obtained by adding

probabilities in a fairly transparent way to “algebraic” (i. e., non-probabilistic) grammars.

• Stochastic grammars supplement underlying algebraic grammars.

Dependencies• Local Dependency: dependence

between two words expressed within the same syntactic rule. (n-grams model this well)

• Non-local dependency: is an instance in which two words can be syntactically dependent even though they occur far apart in a sentence.

Ambiguities1. “Children eat sweet candy”2. “Too much boiling will candy the

molasses”• In sentence (1) candy is a noun

while in (2) it is an adjective.• Word category (POS) ambiguity

needs to be resolved.

Ambiguities (Cont.)• Semantic Roles: Determining thematic roles in a

sentence.• Agent, Patient, Experiencer, Instrument, Goal ….• Raju(AGENT) hit us (PATIENT) with a ball (INSTRUMENT). • Complicated by the notions of direct and indirect

object, active and passive voice.

Ambiguities (Cont.)• Attachment ambiguities occur with phrases

that could have been generated by two different nodes in the parse tree. E.g.: saw the man in the house with a pole.

• Rare Usage and spurious usage: A hectare is a hundred ares.

Garden-Path Sentences• Garden-Path sentences are

sentences that lead you along a path that suddenly turns out not to work. E.g.: The horse raced past the barn fell.

Local and Non-Local Dependencies

• A local dependency is a dependency between two words expressed within the same syntactic rule.

• A non-local dependency is an instance in which two words can be syntactically dependent even though they occur far apart in a sentence (e.g., subject-verb agreement; long-distance dependencies such as wh-extraction).

• Non-local phenomena are a challenge for certain statistical NLP approaches (e.g., n-grams) that model local dependencies.

51

The Place of Syntax• Between Morphology and Meaning• Morphology provides/expects:

– lemmas (now it’s time to extract syntactic information from a dictionary)

– tags (Part-of-Speech and combination of morphological categories, such as number, case, tense, voice, ...)

– and of course, we also have word order now to look at/provide

• Typically multiple input (non-disambiguated morphology) / output (multiple syntactic structures, non-disambiguated)

52

Words, Phrases, Clauses, Sentences

• Words– smallest units on the syntax level

• function/autosemantic• Phrases

– consist of words and/or phrases; “constituents”• Clauses

– have predicative meaning (single predicate)• Sentences

– consist of clauses (one or more)

53

Words• Words

– lexical units• auxiliary (function) words: have grammatical function • autosemantic words (“lexical” words)

– idioms• fixed phrases (non-compositional) -> “words”

• Relate to other words– dictionary: repository of information for each words about its (idio

syncratic) relations to other words

54

Phrases• Phrases

– sequences of words and/or phrases (i.e. of constituents)• may be discontinuous, sometimes

• Types of Phrases:– Simple/Clausal (i.e. clauses, which consist of phrases,

behave like phrases... recursively!)– According to head type:

• Noun: a new book• Adjective: brand new• Adverbial: so much• Prepositional: in a class• Verb: catch a ball

55

Noun Phrases• Head: noun

– water– a book– new ideas– that small village– The greatest rise of interest rates since W.W.II within a

single year– an operating system which, despite great efforts on the

part of our administrators, fails all too often

56

Adjective Phrases• Head: adjective• Simple APs very common, complex APs rare

– old– very old– really very old– five times older than the oldest elephant in our

ZOO– (was) sure, as far as I know, to be there first

57

Adverbial and Numerical Phrases

• Head: adverb– three times as much– quickly– really– (... speaks) more loudly than anybody could

imagine– yesterday

• Numerical Phrases– (... lasted) three hours– twenty-two

58

Prepositional Phrases• Head: preposition• In fact, play the role of Adverbial Phrases often

– in the City– at five o’clock– to a brightest future– without a glitch– to the point where neither of them could get out of

it– up to five points– instead of Charles

59

Verb Phrases

• Head: verb– (It) rains– ... could ever see a large Unidentified Flying

Object– ..., why (we) have got so much rain– Please!– On Sunday, (he) was driven to the hospital– (It) began to snow– (...) prohibits smoking in this area

60

Coordination of Phrases• “Head”: conjunction, punctuation

– and, or, but• cats and dogs• new or even newer• quickly and precisely• he came to the conclusion that it makes no

sense to hide himself anymore and therefore we could hear him today

• (trains) from and to Baltimore• eat your lunch now or at the picnic table

61

Ellipsis• Word or Phrase missing where one would normally expect one;

often happens in dialogues– Whom did you see there?– Peter. ?? verb ??

• Most common in coordination (written text)– Pittsburgh leads 4-0 but Detroit only 3-1. ??verb in 2nd

part?? • Systematic in many languages: pro-drop (leave out a pers.

pronoun in the Subject position)– [She] Passed the exam easily.

62

Clauses• Predicative function:

– some activity of some subjects/objects, somewhere in time, under certain circumstances

• Main clause– not part of a greater clause

• Embedded clause– part of other clause, having some function (like a

phrase)• Function of a Clause

– same as for phrase, plus some (direct speech/discourse etc.)

63

Gaps (Non-Continuous Constituents)

• Constituent moves from the expected position:– happens in questions and relative clauses

• Who(m) do you work for <gap>whom?– strictly speaking, do you work should be you (do work)

• I don’t know why we have got so much rain <gap>why?

• On Sundays, I usually work <gap>On Sundays but I stay home on Tuesdays.

• The story he never wrote <gap>the story

• And finally the car she was supposed to use <gap>the car for her trip to New York broke.

– The last two: also could be considered ellipsis (which) plus a gap.

64

Sentences• Consist of a single or several main clauses• If several main clauses:

– coordination, much like coordinated phrases– more coordinating conjunctions:

• and, or, but, (and) therefore, ...• In written text, starts with a capital letter• Ends by period/question mark/exclamation mark

• not all periods end a sentence!• Sometimes even semicolon (;) might be a sentence

break (...vague)

65

Syntax: Representation• Tree structure (“tree” in the sense of graph theory)

– one tree per sentence• Two main ideas for the shape of the tree:

– phrase structure (~ derivation tree, cf. parsing later)• using bracketed grouping• brackets annotated by phrase type• heads (often) explicitly marked

– dependency structure (lexical relations “local”, functions)• basic relation: head (governor) - dependent• links (edges) annotated by syntactic function (Sb, Obj, ...)• phrase structure: implicitly present (but 1:n mapping Dep→PS)

66

Phrase Structure Tree• Example:

((DaimlerChrysler’s shares)NP (rose (three eights)NUMP (to 22)PP-NUM )VP )S

67

Dependency Tree• Example:

rosePred(sharesSb(DaimlerChrysler’sAtr),eightsAdv(threeAtr),toAuxP(22Adv))

Semantic Roles

• Most commonly, noun phrases are arguments of verbs. These arguments have semantic roles: the agent of an action, the patient and other roles such as the instrument or the goal.

• In English, these semantic roles correspond to the notions of subject and object.

• But things are complicated by the notions of direct and indirect object, active and passive voice.

Subcategorization• Different verbs can relate different numbers of entitie

s: transitive versus intransitive verbs.• Tightly related verb arguments are called complemen

ts but less tightly related ones are called adjuncts. Prototypical examples of adjuncts tell us time, place, or manner of the action or state described by the verb.

• Verbs are classified according to the type of complements they permit. This called subcategorization. Subcategorizations allow to capture syntactic as well as semantic regularities.

Attachment Ambiguity and Garden-Path Sentences

• Attachment ambiguities occur with phrases that could have been generated by two different nodes in the parse tree.The child ate the cake with a spoon.

• Genuinely ambiguous: Fruit flies like a banana.• Garden-Path sentences are sentences that lead along

a path that suddenly turns out not to work.The horse raced past the barn fell.

Semantics• Semantics is the study of the meaning of words, constructions,

and utterances.• Semantics can be divided into two parts: lexical semantics and

combination semantics.• Lexical semantics: hypernymy, hyponymy, antonymy, merony

my, holonymy, synonymy, homonymy, polysemy, and homophony.

• Compositionality: the meaning of the whole often differs from the meaning of the parts.

• Idioms correspond to cases where the compound phrase means something completely different from its parts.

Pragmatics

• Pragmatics is the area of studies that goes beyond the study of the meaning of a sentence and tries to explain what the speaker really is expressing.

• Understand the scope of quantifiers, speech acts, discourse analysis, anaphoric relations.

• The resolution of anaphoric relations is crucial to the task of information extraction.