Methodologies, Tools and Languages. Where is the Meeting Point? Asunción Gómez-Pérez Mariano Fernández-López Oscar Corcho Artificial Intelligence Laboratory Technical University of Madrid (UPM) Spain
Index Methodologies Technology Languages
Methodologies 1. Presenting the most representative methodologies - Uschold & King - Grüninger & Fox - Bernaras et alia - Methontology - Sensus 2. Performing an analysis of such methodologies against the same framework of reference 3. Out of this presentation: ontology learning methods
Uschold s Methodology 1. Identify Purpose and Scope 2. Building the ontology Capture Coding Integrating existing ontologies Identify key concepts and relationships Produce unambiguous text definitions Identify terms to refer to such concepts and relations Commit to a meta-ontology Choose a representation language Write the code How and whether to reuse ontologies that already exist 3. Evaluation 4. Documentation 5. Guideliness for each phase Uschold, M.; Grüninger, M. ONTOLOGIES: Principles, Methods and Applications. Knowledge Engineering Review. Vol. 11; N. 2; June 1996.
TOVE Methodology Motivating Scenarios Informal Competency Questions Formal Terminology Formal CQ Formal Axioms Completeness Theorems Identify intuitively possible applications and solutions As an entailment of consistency problems with respect to the axioms in the ontology Conditions under which the solutions to the questions are complete Identify Queries: Answers: Axioms Formal definitions Questions: Terminology Defined as a first-order sentence using the predicates of the ontology Objects Attributes Relations KIF Constants Variables Functions Predicates Uschold, M.; Grüninger, M. ONTOLOGIES: Principles, Methods and Applications. Knowledge Engineering Review. Vol. 11; N. 2; June 1996.
Methodology used on the KACTUS project A botton-up approach for building ontologies Build a preliminary ontology for refinement and augment with new definitions Specification of the application Preliminary design based on relevant top-level ontological categories Redefine Domain A. Bernaras; I. Laresgoiti; J. Corera. Building and reusing ontologies for electrical network applications ECAI96. 12th European Conference on Artificial Intelligence. 1996. 298-302
SENSUS as a basis for a domain-specific ontology Top-down approach Linking Domain Specific Terms to a broad Coverage To identify the terms in SENSUS that are relevant to a particular domain and then prune the skeletal ontology using heuristics SENSUS SENSUS Skeletal B. Swartout; R. Patil; k. Knight; T. Russ. Toward Distributed Use of Large-Scale Ontologies Ontological Engineering. AAAI-97 Spring Symposium Series. 1997. 138-148.
METHONTOLOGY Framework ONTOLOGY Can be public Define- (Imported ontologies...) The world of ontologies METHODOLOGY Item 1: It is necessary. Item 2: Since Tools To set up a life cycle Development process Gómez-Pérez, A. Knowledge Sharing and Reuse. In the Handbook of Applied Expert Systems. CRC Press. 1998.
Planification Life Cycle Control Quality control Management activities Technical activities Specification Conceptualization Formalization Implementation Maintenance Support activities Acquisition Integration Evaluation Documentation Configuration Management Fernández-López, M.; Gómez-Pérez, A.; Rojas M.D. s Crossed Life Cycle. Lectures Notes in Artificial Intelligence Nº 1937. October 2000
Inter-dependencies Inter-dependencies refer the relationship between activities carried out when building different ontologies O 1 O 2 O 3 Fernández-López, M.; Gómez-Pérez, A.; Rojas M.D. s Crossed Life Cycle. Lectures Notes in Artificial Intelligence Nº 1937. October 2000
Monatomic Ions : Life Cycle Knowledge Acquisition Specification Find and Locate Ontologies Inspect Content and Granularity Selection of Standard-Units and Chemical-Elements Preliminary Evaluation of Standard Units at Ontolingua S. No taxonomic organization Preliminary Evaluation of Chemicals at ODE Several versions
Monatomic Ions : Life Cycle Specification Reuse of Chemical and Standard Units Knowledge Acquisition Conceptualization Integration Revision of Chemical Elements Merge Evaluation C. Management Standard Units Ontological Reengineering
Monatomic Ions : Life Cycle Specification Reuse of Chemical and Standard Units Knowledge Acquisition Conceptualization Integration Revision of Chemical Elements Merge Evaluation C. Management Reverse Engineering Evaluation Redesing Standard Units Ontological Reengineering C. Management Evaluation Implementation
Monatomic Ions : Life Cycle Specification Knowledge Acquisition Conceptualization Integration Revision of Chemical Elements Merge Evaluation C. Management Reverse Engineering Evaluation Redesing Standard Units Ontological Reengineering C. Management Evaluation Implementation Implementation Evaluation Documentation
s crossed life cycles CHEMICAL-ELEMENTS Version 1 Evaluation of v.1 Development ODE Version 2 Version 3 Evaluation of v.2 Evaluation of v.3 Merge + Evaluation + Configuration management Phases SPECIFICATION CONCEPTUALIZATION IMPLEMENTATION MONATOMIC IONS ODE Intradependencies Specification Acquisition Evaluation Documentation Conceptualization Acquisition Evaluation Documentation Integration Implementation Acquisition Evaluation Documentation Integration STANDARD UNITS ODE Ontolingua Server Reengineering + Configuration management Development Evaluation Maintenance Maintenance of Stanford version
Methodology compliance with IEEE Standard. Uschold & King Management processes Pre-develop. Development-oriented processes Development Design Requirements Implementation Post -develop. Integral processes Grüninger & Fox Bernaras et alia Methontol. Sensus
Maturity of the Methodologies Recomended life cycle Compliance with IEEE Std. Recommended techniques Ontologies and applications Detail of the methodology Uschold & King Not known 1 domain Very little Grüninger & Fox Not known 1 domain Little Bernaras et alia Not known 1 domain Very little Methontol. Sensus Not known Several domains Several domains A lot Medium
Conclusions about methodologies 1. None of the methodologies are fully mature if we compare them with the IEEE Standard - METHONTOLOGY - Grüninger & Fox - Uschold & King - SENSUS - Bernaras et alia Maturity degree 2. The proposals are not unified 3. Methodologies should be supported by tools.
Index Methodologies Technology Languages
Technological dimension. Several ontology editors Ontolingua (KSL, Stanford) Ontosaurus (ISI) OILed (University of Manchester) Some ontology-based services : Merge: Chiamera, OntoMorph, ProtégéPromp Translation: Protégé2000, WebODE,... Machine learning (knowledge acquisition) OntoEdit (Ontoprice, Karlsrhue Univ) Protégé2000 (SMI, Stanford) WebOnto (KMI, Open University) WebODE (UPM)
Main Dimensions of this comparative study General description features (developers, releases, current uses, etc.) Sw architecture and tool evolution Interoperability (with other tools & information systems, translations, etc.) Underlying knowledge model Libraries Inference services Methodology support Cooperation Usability aspects: Help system, edition & visualization
General description of tools EU tools USA tools
Architecture and Evolution We are moving into a new generation of Java based Tools WebOnto, OILed, OntoEdit, WebODE, Protégé2000 Just few tools using databases for storing ontologies: Protégé2000, WebODE and OntoEdit (comercial) Only backup management in WebODE Extensibility facilities in OntoEdit, WebODE and Protégé-2000
Knowledge Model & Inference Engine Knowledge Model Most of the tools have the same expressiveness Metaclasses are allowed on Ontosaurus and Protégé-2000 Most of the tools allow to represent axioms INFERENCE ENGINE Ontolingua and Protégé do not have an inference engine OntoEdit, WebODe, Ontosaurus provide evaluation facilities Only DL language (OILed and Ontosaurus) allows automatic classifcations No exception handling
Interoperability New tools export and import to ad-hoc XML and other markup languages But..., what is the quality of all these exports? what about the interoperability between tools? Methodology Support Only WebODE gives support to a methodology. No project management facilities, no (semi)automatic acquisition facilities, no maintenance and a little suport for verification Conceptualization is only included in Ontoedit, Protégé2000 and WebODE Cooperation WebOnto has the most advanced features on collaborative construction Tools need more features to ensure a successful collaborative building of ontologies All help systems must be improved
Technological dimension. Problems No interoperability between tools Different knowledge models Different technology Difficult integration No specialized modules for evaluation, configuration management, automatic construction of ontologies, upgrades of ontologies,... Translation services between tools OntoEdit Protégé2000 Translation services WebODE OILEd
-Based Applications Towards an workbench for ontologies Semantic Portals Middleware Metrics services Brokers... access services Knowledge Management selection services Development Suite Administration services library Query services Ontologies... Component-based Easy integration RAD... editor merge translation evaluation conf. man. acquisition browser mapper docum. evolution Development Tools
WebODE Workbench Status Middleware access services selection services Administration services library Query services... Ontologies editor merge translation evaluation conf. man. done acquisition browser mapper docum. evolution In progress Development Tools
Index Methodologies Technology Languages
Motivation (I) I must develop an ontology. What language do I use to code it??? The one(s) I like the most? The one(s) I know the best? The one(s) supported by an ontology tool? The one(s) that best fit(s) my needs? Corcho, O; Gómez-Pérez, A.; A RoadMap on Specification Languages Lectures Notes in Artificial Intelligence Nº 1937. October 2000
Motivation (II) Most decisions are based on the preferences of the developer But they should be taken on the basis of... - the needs of expressiveness - the needs of reasoning features -... of the application which will use the ontology A deep study of existing languages is needed in order to avoid blind decisions
Prototype of e-commerce ontology Ontolingua, OKBC, FLogic, LOOM, OCML,... Comparison & Assessment RDF(S), OIL, XOL, OML/CKML,... Evaluation Framework (EKAW-00, Corcho and Gomez-Perez)
Evaluation Framework Classes: KR (Expressiveness) Metaclasses Attributes Facets Inference mechanisms (Reasoning) Exceptions Automatic Classifications Taxonomies and partitions Procedures Relations/Functions Instances/Individuals/ Facts/Claims Axioms Production Rules Inheritance Monotonic, non-monotonic Simple, Multiple Execution of procedures Constraint Checking Reasoning with rules
Specification Languages (I) Traditional ontology languages Ontolingua/KIF LOOM OKBC OCML Flogic Standards & Recommendations of W3C XML RDF(S) specification languages SHOE XOL OML/CKML OIL DAML+OIL OIL DAML+OIL XOL SHOE OML RDF(S) XML
Formalisms and languages Language Ontolingua/KIF OKBC OCML LOOM FLogic SHOE XOL OIL DAML+OIL OML/CKML RDF(S) Formalism Frames First order Logic Description Logic Conceptual Graphs Semantic Nets
Comparison & Assessment (II) Assessment: Which language should I choose??? High expressiveness needs Traditional languages Automatic Classifications Description logic exchange Web-based languages Agent-based architectures exchange Reasoning Web-based languages Traditional languages
Conclusion Methodologies, tools and languages Methodologies Tools Languages Uschold y King Grüninger y Fox Bernaras et alia SENSUS METHONTOLOGY Oiled OntoEdit WebODE Protégé-2000 Ontolingua Server Ontosaurus WebOnto DAML+OIL OIL RDF (S) RDF XML KIF Ontolingua OKBC LOOM OCML Flogic CARIN
Methodologies, Tools and Languages. Where is the Meeting Point? Asunción Gómez-Pérez Mariano Fernández-López Oscar Corcho Artificial Intelligence Laboratory Technical University of Madrid (UPM) Spain