controlled ontology evolution through semiotic-based ontology evaluation international workshop on...
TRANSCRIPT
Controlled Ontology Evolution Through Semiotic-based Ontology Evaluation
International Workshop on Ontology Dynamics IWOD 2008
Renata [email protected]
Daniel [email protected]
Overview
Motivation Semiotic-based Ontology Evaluation Controlled Ontology Evolution through Evaluation Semiotic-based Ontology Evaluation Tool Evaluation Results Conclusion
Motivation = Reason for changes!
not good for my application?Apply changes!
not good for my intend of
use?Apply changes!
Inconsistent?Apply changes!
not aligned to dependent ontologies?
Apply changes!
Motivation = Evaluation criterion!
Which version is the best one
for my app?
Is the ontology still good (or better) for my intend of use?
Is the ontology consistent?
Can I apply these changes
without affecting the dependent
ontologies?
Overview
Motivation Semiotic-based Ontology Evaluation
Semiotics Semiotics & Ontology Semiotics & Ontology Evaluation
Controlled Ontology Evolution through Evaluation Semiotic-based Ontology Evaluation Tool Evaluation Results Conclusion
Yojo
Concept
SymbolObject
The Meaning Triangle: (Ogden and Richards 1923)
Semiotics is composed of three fundamental
components: (Moris, 1938)
Syntax
Semantics
Pragmatics
(Niles & Pease, 2001)
Language
Syntax
Semantics
Pragmatics
OntologySemiotic Object
Semantics
Pragmatics
Syntax
SymbolObject
Concept
SymbolOntology Graph
graph-like structures containing terms and their
inter-relationships
ObjectIntended Conceptualization
represent an intended conceptualization.
ConceptCommunication Context
Rep
The ontology’s representation is interpretable by some agent
(Gangemi at al, 2006)
Structural Measures
DepthBreathModularity …
Syntax
(Topological Dimension)
ConsistencyComplexityConcept Satisfiability …
Formal Semantics
(Logical Dimension)
Semiotic Measures
Pragmatics Measures
Expert‘s judgmentsUser satisfactionAgreement satisfaction
Data SetTask AssessmentModularity Assessment…
Precision-Recall Based Measures
(Functional Dimension)
FunctionalMeasures
Assesing the ontology pragmatics
Assesing the ontology cognitive semantics
Assesing the ontology syntax and formal semantics
Annotations/Documentation:Structural-Related: Depth, Breath, Modularity.Functional-Related: Expert‘s Judgments, Data Set.User-Oriented: Deployment, Commercial , History/Review, Version.
Ontology profile (Usability Dimension)
Overview
Motivation Semiotic-based Ontology Evaluation Controlled Ontology Evolution through Evaluation Semiotic-based Ontology Evaluation Tool Evaluation Results Conclusion
Controlled Ontology Evolution through Evaluation
Semiotic Ontology Evaluation Process
Capturing RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
(Stojanovic, 2004)
Controlled Ontology Evolution through Evaluation
Semiotic Ontology Evaluation Process
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
Explicit RequirementsReasons for changes = Evaluation Criteria
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
Making the ontology changes visible in a form of an adequate representation
Ontology changes need to be managed such that the ontology remains consistent
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
FunctionalEvaluation
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
FunctionalEvaluation
Ontology changes need to be managed such that the ontology remains consistent
Verify consistency of dependent ontologies
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
FunctionalEvaluation
PragmaticsEvaluation
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
FunctionalEvaluation
PragmaticsEvaluation
User is able to approve the changes applied or to reverse them
Controlled Ontology Evolution through Evaluation
Capturing
RepresentationSemantics of
changePropagation Implementation Validation
Controlled Ontology Evolution through Evaluation
StructuralEvaluation
FunctionalEvaluation
PragmaticsEvaluation
Overview
Motivation Semiotic-based Ontology Evaluation Controlled Ontology Evolution through Evaluation Semiotic-based Ontology Evaluation Tool
Implementation
Evaluation Results Conclusion
Semiotic-based Ontology Evaluation Tool
StructuralEvaluation
FunctionalEvaluation
PragmaticsEvaluation
Semiotic-based Ontology Evaluation Tool
RepresentationSemantics of
changePropagation Implementation Validation
Semiotic Measures
Consistency Checking:
onto changes remains consistent
Task-based Approach:
onto. changes max. performance
Annotation Analysis:
changes reported for versioning
Implementation:
Semiotic-based Ontology Evaluation Tool
Validation
Semantics ofchange
Propagation
RepresentationSemantics of
change
Use of the reasoners RACER System and Pellet.
„…a logical theory is consistent if it does not contain a contradiction, or, more precisely, for no proposition φ are both φ and ¬φ provable.“
Person Seal
Disjoint (Person, Seal)
Shark (primitive class)Animal andeats some (Person and Seal)
Inconsistent
Consistency Checking
Task-based Approach
How effective a given ontology is in the light of a well-defined task (Porzel, 2004; Maedche & Staab,2002)
Task
Application
Performance Results
Ontology
Compare with Gold Standard
AnswersImprovements ?
Task-based Approach
Evaluate different ontology versions!
OntologyV0.1
OntologyV0.2
changes
Is my evolved ontology still good (or better) for my
intend of use?
max. performance for a specific task!
Task-based Approach
How efficient is the system to answers questions using just ontologies
Improvements
Question-Answering
SmartWeb
Performance Results
Compare to Gold Standard
Answers
SWIntOV0.2
V0.1
Task-based Approach
1. Plug the Evolved Ontology
4. Compare with GS Answers
2. Query the system
3. Check Time Performance
5. Make Report
• Lexicon• Taxonomy• Semantic Relations
SmartWeb
Performance Results
Compare withGold Standard
Answers
Evolved
Improvements6. Apply changes!
Usability-Related Evaluation
Annotation Analysis: Quantitative analysis of the amount of metadata linked to the tag ”rdf:comments”
<owl:AnnotationProperty rdf:ID=“structural“/><owl:AnnotationProperty rdf:ID=“functional“/><owl:AnnotationProperty rdf:ID=“user-oriented“/>
<owl:AnnotationProperty rdf:ID=“consistency“> <rdfs:subPropertyOf rdf:resource=“#structural“/></owl:AnnotationProperty><owl:AnnotationProperty rdf:ID=“task-assessment“> <rdfs:subPropertyOf rdf:resource=“#functional“/></owl:AnnotationProperty>…
<owl:Class rdf:ID="Teacher"> <rdfs:comment>Teacher Class </rdfs:comment> <rdfs:subClassOf rdf:resource="#Person"/> </owl:Class>
Overview
Motivation Semiotic-based Ontology Evaluation Controlled Ontology Evolution through Evaluation Semiotic-based Ontology Evaluation Tool Evaluation Results Conclusion
Evaluation Results
SWIntO Ontology (SmartWeb Project*)
Foundational (DOLCE) and general (SUMO) knowledge Domain- and task-specific knowledge
Football (soccer) entities and events Navigation Linguistic information Discourse Multimedia
* http://www.smartweb-project.org/
SmartDOLCE:Entity
SmartSUMO:Attribute
SmartSUMO:SocialRole
SportEvent:FootballPlayer SportEvent:FootballOrganizationPerson
…
… …
……
……
Consistency Checking
Functional Evaluation I
SWIntO V.0.3.2
Q1:Which matches took place in the semifinals in 1954?
Q2:Who was the world champion in 1990?
Time-performance:31,10 msGS-performance:26,23 ms
Vocabulary Overlap = 100%Hierarchy Overlap = 87%Relation Overlap = 45%
Evaluated Relation:
GS Relation:
SWIntO V.0.3.3
Q1:Which matches took place in the semifinals in 1954?
Q2:Who was the world champion in 1990?
Time-performance:31,10 msGS-performance:26,23 ms
Vocabulary Overlap = 100%Hierarchy Overlap = 87%Relation Overlap = 45%
List of Overlap Descriptions:
Evaluated Relation:
GS Relation:
Functional Evaluation II
Annotation Analysis
Conclusion
Evaluation framework to support and control ontology evolution Apply changes to an ontology keeping its quality with respect to the purpose
of the ontology (or the purpose of the ontology changes) Controlled evolution by assessing the quality of the ontology with
respect to all semiotic dimensions -> Ontology changes captured by ontology evaluation process
Implementation = choose three measures which are essential in any ontology evolution/evaluation process
Structural Dimension: Consistency Checking Functional Dimension: Task-based Evaluation Usability Dimension: Annotation Analysis
Future Work level of granularity & integration
Thank you for your attention!
RACER System, Renamed abox and concept expression reasoner . http://www.racer-system.comR. Porzel and R. Malaka, A task-based approach for ontology evaluation, 2004.M. Ciaramita J. Lehmann A. Gangemi, C. Catenacci, Modelling ontology evaluation and validation.V. Sugumaran A. Burton-Jones, V. C. Storey and P. Ahluwalia, A semiotic metrics suite for assessing the quality of
ontologies, 2004.Janez Brank, Marko Grobelnik and Dunja Mladenic, A Survey of Ontology Evaluation Techniques.W. Wahlster, Smartweb: Mobile applications of the semantic web, In Proceedings of Informatik. 2004s. P. Buitelaar and A. Frank,Ontology-driven Predicate-Argument Structure Analysis for Event AnnotationJohn F. Sowa, Ontology, Metadata, and Semiotics, 2000.P. Cimiano,Text Analysis and Ontologies, 2006.C.W. Morris, Foundations of a theory of signs. In: International Encyclopedia of Unified Science (O. Neurath, R. Carnap & C.
Morris, eds), Chicago University Press, Chicago, pp. 77-138. 1938.I. Niles and A. Pease .Towards a Standard Upper Ontology, 2001.Y. Sure and R. Studer. On-to-knowledgemethodology - final version. Technical Report Deliverable 18, Institute AIFB,
University of Karlsruhe, 2002.A. Gangemi, C. Catenacci, M. Ciaramita, and J. Lehmann. Qood grid: A metaontology-based framework for ontology
evaluation and selection. In Proceedings of EON2006, 2006.N. Guarino.. Towards a Formal Evaluation of Ontology Quality, pages 1541–1672. IEEE Intelligent Systems, 2004.D.Orbele et al. Dolce ergo sumo: On foundational and domain models in swinto (smartweb integrated ontology), 2006..N. Noy. Evaluation by ontology consumers. pages 1541–1672. IEEE Intelligent Systems, 2004.K. Dellschaft and S. Staab. On How to Perform a Gold Standard Based Evaluation of Ontology Learning, 2006A. Maedche and S. Staab.Measuring Similarity between Ontologies, 2002.L. Stojanovic. Methods and tools for ontology evolution. PhD thesis, University of Karlsruhe (TH), 2004.
References