18
Rencontres sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011 Tizi-Ouzou, Algérie Empty Answers to Flexible Queries: A Novel Approach AGGOUNE Aïcha Presented by: Aicha AGGOUNE 1 , Allel HADJALI 2 1 University of Sétif, Algeria 2 IRISA/ENSSAT, Université de Rennes 1, France

Rencontres sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011 Tizi-Ouzou, Algérie

  • Upload
    tallis

  • View
    32

  • Download
    2

Embed Size (px)

DESCRIPTION

Rencontres sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011 Tizi-Ouzou, Algérie. Empty Answers to Flexible Queries: A Novel Approach. Aicha AGGOUNE 1 , Allel HADJALI 2 1 University of Sétif, Algeria 2 IRISA/ENSSAT, Université de Rennes 1, France . Presented by: . - PowerPoint PPT Presentation

Citation preview

Page 1: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Rencontres sur la Recherche en Informatique R2I 201112-14 Juin 2011

Tizi-Ouzou, Algérie

Empty Answers to Flexible Queries: A Novel Approach

AGGOUNE AïchaPresented by:

Aicha AGGOUNE1, Allel HADJALI2 1University of Sétif, Algeria

2IRISA/ENSSAT, Université de Rennes 1, France

Page 2: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Content

Introduction & Problematic1

Flexible queries & EAP2

3Measuring proximity

4 Approaching the EAP

Conclusion & perspectives5

Page 3: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Introduction & Problematic

One of the key elements of the quality of DBMS is the pertinence of the answers, in particular, by the consideration of the preferences of the users in the queries.

It  was  shown  that the  theory of  fuzzy  sets  provides  effective tools for expressing preferences. In this case, queries are called the "Flexible queries (or Fuzzy) ".

One of the problems with which the users are confronted, in a process of interrogation, is the problem of the empty answers (EAP).

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie  02/16

Page 4: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Introduction & Problematic

The flexible queries limit to a large extent the risk of obtaining of the empty answers, however this risk is not totally eliminated.

?Proposition of a semantic proximity measure between flexible queries for the treatment of the EAP, this measure uses a particular distance, called the Hausdorff distance.

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   03/16

Page 5: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Flexible queries & EAP

Generalization of boolean query

Included the vague and imprecise terms

Expression of criteria

importance levels

Expression of preferences

Flexible Query

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 04/16

Page 6: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Flexible queries & EAP

Gradual Operators (Clearly superior to)

Simples fuzzy predicates (Cheap, Expensive,…)

Fuzzy predicates + modifiers (Very, relatively,…)

Compund fuzzy predicates (Young andWell Paid,…)

Fuzzy predicates + quantifiers

(plus, less, …)

Fuzzy Criteria

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 05/16

Page 7: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Flexible queries & EAP

Flexible Query Languages :

Extensions of classical interrogations languages .

The result of a flexible query is a fuzzy relation for which every tuple t is a degree of membership noted / t.

Extension of the relational algebra. Ex: rUs(u)= max(r(u), s(u))

Three languages of type SQL: SQLF, SQLF Etendu, FQUERY

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   06/16

Page 8: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Example: (SQLF)

SELECT * FROM VOITURE

WHERE Marque = ‘Famous' AND Prix = ‘Cheap';Mark Colour Price

BMW vert 30000 AUDI rouge 15100VW blanc 15000

Mark Colour Price Degree Mark Degree Price

BMW vert 30000 0.9 0.2AUDI rouge 15100 0.7 0.7VW blanc 15000 0.5 0.9

Mark Colour Price Degree

BMW vert 30000 0.2AUDI rouge 15100 0.7VW blanc 15000 0.5

Flexible queries & EAP

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   07/16

Page 9: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Flexible queries & EAP

«  The problem of empty answers means that all the answers does not inform the user about their questions and satisfy their needs. "

Godfrey  «A query is said an empty response if it produces no result»  Motro

Empty Answers Problem : EAP

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 08/16

Page 10: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Hausdorff Distance

The Hausdorff distance is a topological tool to deduct the similarity between fuzzy sets, it measures the resemblance or the dissimilarity of two subsets of a metric space.

The Hausdorff distance between two sets A and B is : dH(A, B) = max {supaA infbB d(a, b), supbB infaA d(a, b)}

In the case of a discrete fuzzy set and according to Rosenfeld‘s approach

Measuring proximity

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 09/16

Page 11: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

In the continuous case :

The sementic proximity is: . The simplest form is :  

Ranking  approximate queries is done by  descending order of semantic proximity and number of approximate answers.

Measuring proximity

Hausdorff Distance

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   10/16

Page 12: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Single Predicate Queries:Let Q=P and Q'=P' (P, P' are pertaining to the same attribute A ) :

Compound Queries (more predicates ):Let Q= P1P2…Pk

There are 3 cases: Case 1: Q and Q' cover the same attributes exactly. Case 2: Q' covers all the attributes specified in Q. Case 3: Q' does not cover all the attributes specified in Q.

Q'= P'1 P'2… P'k (Pi, Pi' are pertaining to the same attribute Ai) 

Case 2: Q' covers all the attributes specified in Q.

Q'=P'1 ... P'kP'k+1… P's.

(S1) Increased Q : Q = P1 ... Pk D(Ak+1 ) … D(As).

(S2) Weakening of Q' : Q'R = P'1 ... P'k .

Case 3: Q' does not cover all the attributes specified in Q. Q' = P'1 ... P'b P'k+1 … P's with b < k and (Pi, P'i) D(Ai), for i = 1 to b, predicats Pj (for j = k+1 to s) are not specified in Q.

i. Increased Q' : Q' = P'1 ... P'b D(Ab+1)… D(Ak) P'k+1 … P's

ii. Applying one of two strategies described in case 2.

Measuring proximity

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 11/16

Page 13: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Approaching the EAP

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   12/16

Using a set of queries previously evaluated by the system and whose answers are not empty (Workload of past queries).Given the workload, estimating the semantic proximity of the empty answers query Q whith queries of the workload.

The approximate answer to the query is a set of tuples ordered by the closest query in query EA.

Principle

Page 14: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Principle

Approaching the EAP

Treatment of EAP

Algorithme TEA

Q

Q

Q'

QappQ

∑Q

∑Qapp

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   13/16

Page 15: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Partitionnons the workload W ( D ) in 03 subsets:W=(D, Q) = {Q' / Q' W(D) |D(Q') D(Q)| = k}, W> (D, Q) = {Q' / Q' W(D) |D(Q') D(Q)| > k},W< (D, Q) = {Q' / Q' W(D) |D(Q') D(Q)| < k},

(a)- If W= (D, Q) , for element Q' W=(D, Q), estimate the proximity Prox(Q, Q'). Ranking in descending order of queries.

Choose the closest query Qapp in Q, and affect it as degree of membership, the value of proximity measured Prox (Q, Qapp).

1

2

3

(b)- Else if W >(D, Q) , then we consider W >(D, Q) and apply (a).

Approaching the EAP

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie   14/16

Approach

(c)- Else, we consider W<(D, Q) and apply (a).

Page 16: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Conclusion & perspectives

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 15/16

The introduction of flexibility on the elements of DB, based on the theory of fuzzy sets;

The proposition of a measure of semantic proximity between flexible queries;

Using this measure to treat the EAP.

The main contributions are:

Page 17: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie

Conclusion & perspectives

R2I'2011, 12-14 Juin 2011, Tizi Ouzou, Algérie 16/16

Experiment of the proposed approach.

Proposition a search algorithm for closest queries to EA query.

Analysis of results.

Nevertheless, our approach opens future perspectives. In the short term, the most important are the following ones:

Compare the proposed approach to existing approaches.

Page 18: Rencontres  sur la Recherche en Informatique R 2 I 2011 12-14 Juin 2011  Tizi-Ouzou,  Algérie