Matching Domain Ontologies A Comparative Study [Mode De Compatibilité]

Preview:

Citation preview

Mrs Leila Mrs Leila Mrs Leila Mrs Leila Mrs Leila Mrs Leila Mrs Leila Mrs Leila GhomariGhomariGhomariGhomariGhomariGhomariGhomariGhomari--------ZemmouchiZemmouchiZemmouchiZemmouchiZemmouchiZemmouchiZemmouchiZemmouchiL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dzL_zemmouchi@esi.dz

JanuaryJanuaryJanuaryJanuaryJanuaryJanuaryJanuaryJanuary, 31, 31, 31, 31, 31, 31, 31, 31st st st st st st st st 20092009200920092009200920092009

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

22

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Context

Need :

OntologyOntologySimultaneousSimultaneous

OntologicalEngineering

DataIntegration

P2P Information sharing

3

SimultaneousSimultaneousUtilizationUtilization

sharing

Web Services Composition

Multi-Agent Communication

1.. Syntactic Level

2. Terminological Level

ContexteContext : Ontology Heterogeneity

3. Conceptual Level

4. Semiotic Level

4

Context : Ontology Heterogeneity

Ontology Matching Ontology Matching

process of corresponding semantically process of corresponding semantically

5

Entities which compose Entities which compose ontologiesontologies

Context : Ontology Heterogeneity

6

Adapted from [Isaac, 2007]

Context : Ontology Heterogeneity

7

Adapted from [Isaac, 2007]

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

88

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Research Problem & Objectives

9

Research Problem & Objectives

Since2004

10

I3CON EON

Information Interpretationand IntegrationConference

Evaluation of ONtology

Tools

OntologyAlignmentEvaluation Initiative

Research Problem & Objectives

Matching Systems

Cupid

TranScm

SKAT

RiMOM

Hovy

MapOnto

Clio

Falcon-AO

H-Match

Artemis

Tess

DIKE

ASCO

Similarity flooding

OLA

Automatch

Wise-Integrator

Anchor-Prompt

OMEN

BayesOWL

OntoMerge

MoA

HCONE

11

Falcon-AO

oMap

ToMAS

XClust

SBI&NB

Kang & Naughton

Wang & al.

NOM & QOM

Dumas

LSD/GLUE/iMAP

FCA-merge

IF-Map

Xu & al.

COMA & COMA++

DCM

T-tree

HCONE

DELTA

sPLMap

SEMINT

CAIMAN

S-Match

OntoBuilder

CtxMatch

Corpus-basedmatching

Research Problem & Objectives

““OntologiesOntologies are formal representations of are formal representations of semanticssemantics””

[[GuarinoGuarino, 1995], 1995]

12

Research Problem & Objectives

Syntaxic Systems

T-tree

Wise-Integrator

Anchor-Prompt

OMEN

BayesOWL

OntoMerge

MoA

Semantic Systems

S-Match

CtxMatch

13

MoA

HCONE

DELTA

sPLMap

SEMINT

CAIMAN

COMA & COMA++

OntoBuilder

OLA

Research Problem & Objectives

14

Research Problem & Objectives

Contribute to analyze the progress of both semanticsyntactic matching systems

1

Identify selected matching systems strengths andweaknesses in order to improve their matching quality.

15

Contribute to analyze the progress of both semanticsyntactic matching systems

2

3

Help future matching systems developers to selectthe adequate approach matching

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

1616

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Ontologies Selection

To Evaluatean

automaticmatchingresult

To Achievea referencematchingwhich ismanual

To Undestandontologies

to bematchedvery well

To beOntologies Domain(s) Experts

17

Ontologies Selection

18

Source : [Sean & al.], OM 2007

Ontologies Selection

Ontology URI University Origin

O1 http://www.mindswap.org/2005/debugging/ontologies/

University.owl

19

University.owl

O2 http://www.lehigh.edu/~zhp2/2004/0401/

univ-bench.owl

O3 http://www.webkursi.lv/luweb05fall/resources/

university.owl

Ontologies Selection

Ontologies Classes Properties Restrictions Instances Language

O1 30 12 18 4 OWL - FULL

20

O2 43 31 8 0 OWL - DL

O3 73 46 33 80 OWL - FULL

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

2121

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Matching Systems Selection

CTXMatch

2003S-Match

2004

(not available)

S-Match

2004

(not available)

22

CTXMatch 2

2006

OWL-CTXMatch

2006

[Bouquet & al., 2006]

Trento University, Italy

Matching Systems Selection

DELTA MapOnto XClustDumas Wang & al. SEMINT

HovyClio

SBI&NBGLUE DCM CAIMAN

Cupid Falcon-AOKang &

NaughtonFCA-merge T-tree QOM

COmbination

of schema MAtching

Approaches

[Aumueller & al., 2005]

Leipzieg U., Germany

23

TranScm oMapCOMA++

IF-MapWise-

IntegratorOntoBuilder

SKATToMAS

ASCO Xu & al. BayesOWL

RiMOM NOMSimilarity

floodingAnchor-Prompt OntoMerge

Corpus-based

matching

Tess H-Match OLA OMEN MoA LSD

DIKE Artemis Automatch sPLMap HCONE IMAP

Matching Systems Selection

24

Matching Systems Selection

SemanticSemanticElicitationElicitation

InternalrepresentationConstruction (Form :description logicformulas)

25

formulas)

AutomaicAutomaicDeductionDeduction of of relationshipsrelationships

betweenbetween entitiesentitiesby a by a reasonerreasoner

The reasoner mergeFormulas sets in onemodel, classify anddetermine whichrelation typeassociates the twoentities (=, ∩,⊆,⊇,⊥)

OWL-CTXMatch

Matching Systems Selection

SchemasSchemasManipulationManipulation

(Entity1, Entity2, Matcher)=

MatchersMatchersDefinitionDefinition and and ExecutionExecution

26

SimilaritySimilarityCubeCube

Matcher)= Similarity value

AgreggationAgreggation

SelectionSelectionCombinationCombination

DirectionDirection

COMA++

Where user can modifythe default configuration

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

2727

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Matching Ontologies

O1

1st Matching3rd Matching

28

O2O3

2nd Matching

Matching Ontologies

29

Matching Ontologies

30

Matching Ontologies

31

Matching Ontologies

32

Matching Ontologies

33

Matching Ontologies

IntuitionIntuition

Lexical Lexical ThesauriThesaurisuchsuch as : as : WordnetWordnet

34

WordnetWordnet

Expert Domain Expert Domain knowledgeknowledge

OntologiesOntologies

Matching Ontologies

ReferenceMatchingO1 O1 �������� O2O2

ReferenceMatchingO2 O2 �� O3O3

ReferenceMatchingO3 O3 �� O1O1

215 Correspondences

662Correspondences

600Correspondences

35

Correspondences Correspondences Correspondences

6.9% of all

correspondences

7.5%of all

correspondences

12%of all

correspondences

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologies selection

Matchingsystemsselection

Matchingontologies :Research

Problem &

3636

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Matching Evaluation

Precision = TP/TP+FP

Recall = TP/TP+FN

FNFNTNTN

EXPERT

37

F-Mesure =

Overall = Recall (2-(1/Precision))

2 * Recall * Precision

Recall + PrecisionTPTP

FPFP

AUTOMATIC

SYSTEM

Matching Evaluation

38

Matching Evaluation

39

Matching Evaluation

40

Matching Evaluation

41

Matching Evaluation

42

Context :

OntologyHeterogeneity

Solution :

OntologyMatching

ContexteProgram Presentation

Ontologyselection

Matchingsystemsselection

Matchingontologies :Research

Problem &

4343

1st Matching

2nd Matching

3rd Matching

Evaluation of

alignments

Synthesis

Of obtainedresults

Problem &Objectives

Synthesis of obtained results

Class Class

44

Class

Property

Class

Property

Synthesis of obtained results

Classe Classe4577

45

Propriété Propriété

32

0.040.08

0.12

Synthesis of obtained results

46

Measuring Unit : second

Synthesis of obtained results

47

Synthesis of obtained results

48

Few Common

alignments

Synthesis of obtained results

Few Common

alignments

49

Few Common

alignments

CONCLUSION

•The two matching dimensions must be takeninto account :

syntacticsyntactic ((MatchingMatching termsterms) ) ANDsemanticsemantic ((MatchingMatching Concepts) Concepts)

About Matching

Approaches

MatchingResults

CONCLUSION

To Draw more

More significantMore significantnumber of Tests

To Draw more

General Conclusions

With regard to

Comparatives Syntactic Systems

Versus Semantic Systems

MoreMoreRecommendationsRecommendationsReferenceReference

CONCLUSION

MoreMoreRecommendationsRecommendations

and and NormsNormsTo To achieveachieve a good a good qualityquality manualmanual

matchingmatching

ReferenceReferenceOntologie(s) Ontologie(s)

ReferenceReferenceAlignmentsAlignments

THE END

Recommended