MDA Standards for Ontology Development

Transcription

1 MDA Standards for Ontology Development Dragan Gašević, Vladan Devedžić, Dragan Djurić University of Belgrade Serbia and Montenegro Introduction Outline Semantic Web and Ontologies Model Driven Architecture: An Overview UML-Based Solutions and Tools OMG s Initiative: Ontology Definition Metamodel and Current Submissions The MDA-Based Infrastructure for Ontology Development 7/5/2004 ICWE2004, Munich, Germany 2 1

2 Introduction Synergy of different research fields (Devedžić, 2002) multidisciplinary approach software patterns, ontologies, databases, Recent technologies extensible Markup languages Semantic Web and ontologies Model Driven Architecture and related standards UML, MOF, XMI developing in parallel, but by different communities! 7/5/2004 ICWE2004, Munich, Germany 3 Introduction A technological space is a working context with a set of associated concepts, body of knowledge, tools, required skills, and possibilities. Although some technological spaces are difficult to define, they can be easily recognized (e.g. MDA, ontologies). In order to get a synergy of different technological spaces we should create bridges between them. Kurtev et al, /5/2004 ICWE2004, Munich, Germany 4 2

3 Introduction Bridging Semantic Web and MDA Semantic Web Model Driven Architecture ontologies Bridges! models 7/5/2004 ICWE2004, Munich, Germany 5 Differences Semantic Web and MDA models models are abstractions they eliminate or simplify concepts and relationship ontologies are intended for knowledge representation if something is known, it should be able to be record in a machine-interpretable manner everything is worth capturing 7/5/2004 ICWE2004, Munich, Germany 6 3

4 Similarities Semantic Web and MDA UML models classes, attributes, relations, inheritance, ontologies classes, properties, relations, inheritance, semantically different, but equivalent! metadata sharing XML! 7/5/2004 ICWE2004, Munich, Germany 7 4

5 Semantic Web and Ontologies What is an ontology? Classic definitions (Gruber, 1993), (Guarino, 1994) a specification of a conceptualization an explicit specification of some topic a formal and declarative representation of some subject area 7/5/2004 ICWE2004, Munich, Germany 2 1

6 What is an ontology? Other important definitions (Swartout and Tate, 1999) the basic structure or armature around which a knowledge base can be built 7/5/2004 ICWE2004, Munich, Germany 3 What is an ontology? Other important definitions (Hendler, 2001) a set of knowledge terms, including the vocabulary, the semantic interconnections, and some simple rules of inference and logic for some particular topic this definition is currently predominant in The Semantic Web community 7/5/2004 ICWE2004, Munich, Germany 4 2

7 What is an ontology? What does an ontology provide? the vocabulary (or names) for referring to the terms in that subject area the logical statements that describe: what the terms are how they are related to each other how they can or cannot be related to each other 7/5/2004 ICWE2004, Munich, Germany 5 What is an ontology? What does an ontology provide? rules for combining terms and relations to define extensions to the vocabulary semantics independent of reader and context a common understanding of topics that can be communicated between users and applications 7/5/2004 ICWE2004, Munich, Germany 6 3

8 What is an ontology? What is the purpose of ontologies? knowledge sharing and reuse description of the concepts and relationships that can exist for an IA or a community of IAs the description is like a formal specification of a program 7/5/2004 ICWE2004, Munich, Germany 7 Examples of Ontologies The Frame ontology (Gruber, 1993) partial vocabulary of the Frame ontology class relation (?relation) class function (?function) class class (?class) relation instance-of (?individual?class) function all-instances (?class) :->?set-of-instances function one-of (@instances) :->?class 7/5/2004 ICWE2004, Munich, Germany 8 4

9 Examples of Ontologies The Frame ontology (Gruber, 1993) partial vocabulary of the Frame ontology relation subclass-of (?child-class?parent-class) relation superclass-of (?parent-class?child-class) relation subrelation-of (?child-relation?parentrelation) relation direct-instance-of (?individual?class) relation direct-subclass-of (?child-class?parent-class) 7/5/2004 ICWE2004, Munich, Germany 9 Examples of Ontologies Petri net ontology (Gašević, 2004) Ontology UML Profile representation (excerpt) 7/5/2004 ICWE2004, Munich, Germany 10 5

10 Examples of Ontologies Petri net ontology (Gašević, 2004) RDFS representation (excerpt) <?xml version="1.0" encoding="iso "?> <!DOCTYPE rdf:rdf [ <!ENTITY rdf ' <!ENTITY a ' <!ENTITY PN_Ontology ' <!ENTITY rdfs ' ]> <rdf:rdf xmlns:rdf="&rdf;" xmlns:a="&a;" xmlns:pn_ontology="&pn_ontology;" xmlns:rdfs="&rdfs;"> <rdfs:class rdf:about="&pn_ontology;absolute Position" rdfs:label="absolute Position"> <rdfs:subclassof rdf:resource="&pn_ontology;position"/> </rdfs:class> <rdfs:class rdf:about="&pn_ontology;arc" rdfs:label="arc"> <rdfs:subclassof rdf:resource="&pn_ontology;modelelement"/> </rdfs:class> <rdfs:class rdf:about="&pn_ontology;feature" rdfs:label="feature"> <rdfs:subclassof rdf:resource="&rdfs;resource"/> </rdfs:class> <rdfs:class rdf:about="&pn_ontology;graphicalfeature" rdfs:label="graphicalfeature"> <rdfs:subclassof rdf:resource="&pn_ontology;feature"/> </rdfs:class> <rdfs:class rdf:about="&pn_ontology;graphicalinformation" rdfs:label="graphicalinformation"> <rdfs:subclassof rdf:resource="&rdfs;resource"/> </rdfs:class> <rdf:property rdf:about="&pn_ontology;id" a:maxcardinality="1" a:mincardinality="1" rdfs:label="id"> <rdfs:domain rdf:resource="&pn_ontology;modelelement"/> <rdfs:domain rdf:resource="&pn_ontology;net"/> <rdfs:range rdf:resource="&rdfs;literal"/> 7/5/2004 </rdf:property> ICWE2004, Munich, Germany 11 Why Ontologies? The role of ontologies in the architecture of The Semantic Web to enable intelligent services information brokers search agents information filters intelligent information integration knowledge management 7/5/2004 ICWE2004, Munich, Germany 12 6

11 Why Ontologies? The role of ontologies in the architecture of The Semantic Web to establish further levels of interoperability (semantic interoperability) on the Web syntactic interoperability: reusability in parsing the data semantic interoperability: mappings between terms within the data, which requires content analysis 7/5/2004 ICWE2004, Munich, Germany 13 Why Ontologies? The role of ontologies in the architecture of The Semantic Web to add a further representation and inference layer on top of the Web s current layers to enable Web-based knowledge processing, sharing, and reuse between applications 7/5/2004 ICWE2004, Munich, Germany 14 7

12 Why Ontologies? Ontologies merely serve to standardize and provide interpretations for Web content To make content machine-understandable, Web pages must contain semantic markup descriptions which use the terminology that one or more ontologies define 7/5/2004 ICWE2004, Munich, Germany 15 Ontology structure The main ontology components (Firidman-Noy & Hafner, 1997) conceptualization taxonomy relations functions axioms Similarity with an OO model (classes, class relations, and objects) 7/5/2004 ICWE2004, Munich, Germany 16 8

13 Tools for Building Ontologies XML / XMLS, RDF / RDFS, and the corresponding development tools Ontology representation languages (The Semantic Web languages) Ontology-development environments (integrated graphical tools) Ontology-learning tools 7/5/2004 ICWE2004, Munich, Germany 17 Semantic Web "Layer Cake" 7/5/2004 ICWE2004, Munich, Germany 18 9

14 Tools for Building Ontologies Ontology representation languages Knowledge Interchange Format (KIF) Simple HTML Ontology Extensions (SHOE ) ISO standard for describing knowledge structures (Topic Maps) DARPA Agent Markup Language (DAML) Ontology Inference Layer (OIL, DAML+ OIL) Web Ontology Language (OWL) 7/5/2004 ICWE2004, Munich, Germany 19 Tools for Building Ontologies The need: a standardized Web ontology language it is emerging at the moment OIL, DAML+ OIL, "OIL, a proposal for such a standard" W3C recommendation OWL ISO efforts 7/5/2004 ICWE2004, Munich, Germany 20 10

15 Ideally, we would like a universal shared knowledge-representation language to support the Semantic Web, but for a variety of pragmatic and technological reasons, this is unachievable in practice. Instead, we will have to live with a multitude of metadata representations. S. Decker et al. Tools for Building Ontologies Ontology Inference Layer (OIL) An OIL processor will also understand RDFS Even simple RDF Schema agents are able to process the OIL ontologies 7/5/2004 ICWE2004, Munich, Germany 22 11

16 Semantic Web "Layer Cake" recommendation recommendation recommendation 7/5/2004 ICWE2004, Munich, Germany 23 Web Ontology Language Good starting point: Roger L. Costello, David B. Jacobs, "A Quick Introduction to OWL Web Ontology Language", presentation, The MITRE Corporation, /5/2004 ICWE2004, Munich, Germany 24 12

17 Web Ontology Language DAML OIL RDF DAML = DARPA Agent Markup Language OIL = Ontology Inference Layer DAML+OIL All were influenced by RDF OWL OWL is now a W3C Recommendation! OWL Lite 7/5/2004 OWL DL 25 OWL Full Web Ontology Language OWL is a set of XML elements and attributes, with well-defined meaning, that are used to define terms and their relationships 7/5/2004 ICWE2004, Munich, Germany 26 13

18 Web Ontology Language OWL extends RDF and RDF Schema Class equivalentproperty sameindividualas... subclassof resource ID... OWL RDF/RDF Schema OWL Vocabulary 7/5/2004 ICWE2004, Munich, Germany 27 Symmetric: if P(x, y) then P(y, x) Transitive: if P(x,y) and P(y,z) then P(x, z) Functional: if P(x,y) and P(x,z) then y=z Web Ontology Language InverseOf: if P1(x,y) then P2(y,x) InverseFunctional: if P(y,x) and P(z,x) then y=z allvaluesfrom: P(x,y) and y=allvaluesfrom(c) somevaluesfrom: P(x,y) and y=somevaluesfrom(c) Summary of OWL Capabilities hasvalue: P(x,y) and y=hasvalue(v) cardinality: cardinality(p) = N mincardinality: mincardinality(p) = N maxcardinality: maxcardinality(p) = N equivalentproperty: P1 = P2 intersectionof: C = intersectionof(c1, C2, ) unionof: C = unionof(c1, C2, ) complementof: C = complementof(c1) oneof: C = one of(v1, v2, ) equivalentclass: C1 = C2 Legend: disjointwith: C1!= C2 Properties: P, P1, P2, etc sameindividualas: I1 = I2 Specific classes: x, y, z differentfrom: I1!= I2 Generic classes: C, C1, C2 Values: v, v1, v2 AllDifferent: I1!= I2, I1!= I3, I2!= I3, Instance documents: I1, I2, I3, etc. Thing: I1, I2, A number: N 7/5/2004 ICWE2004, Munich, Germany P(x,y): property P relates 28x to y 14

19 Web Ontology Language Petri net ontology (Gašević, 2003) OWL representation (excerpt) <?xml version="1.0" encoding="utf-8"?> <rdf:rdf xmlns=" xmlns:rdfs=" xmlns:rdf=" xmlns:owl=" <owl:class rdf:id="modelelement"/> <owl:class rdf:id="structuralelement"> <rdfs:subclassof rdf:resource="#modelelement"/> <rdfs:subclassof> <owl:restriction> <owl:onproperty rdf:resource="#element"/> <owl:mincardinality rdf:datatype=" </owl:restriction> </rdfs:subclassof> </owl:class> <owl:class rdf:id="arc"> <rdfs:subclassof rdf:resource="#modelelement"/> <rdfs:subclassof> <owl:restriction> <owl:onproperty rdf:resource="#multiplicity"/> <owl:maxcardinality rdf:datatype=" </owl:restriction> </rdfs:subclassof> <rdfs:subclassof> <owl:restriction> <owl:onproperty rdf:resource="#fromnode"/> <owl:cardinality rdf:datatype=" </owl:restriction> 7/5/2004 </rdfs:subclassof> ICWE2004, Munich, Germany 29 A Vision The Semantic Web's evolving infrastructure not just a few large, complex, consistent ontologies that great numbers of users share great number of small ontological components consisting largely of pointers to each other Web users will develop them in much the same way that Web content is created ("anarchy") 7/5/2004 ICWE2004, Munich, Germany 30 15

20 Tools for Building Ontologies Ontology-development tools Protégé OntoEdit OilEd Chimaera 7/5/2004 ICWE2004, Munich, Germany 31 Ontology-development environments Protégé Extensible platform (plug-ins) Semantic Web: OWL, DAML+OIL, OIL, Import/Export: OKBC Tab Widget, XML, TX RuleML Tab Widget, Inference & Reasoning: Jess Tab, Algernon Tab, CLISP Tab, Software engineering: UML Storage Backend, XMI Storage Backend, 7/5/2004 ICWE2004, Munich, Germany 32 16

21 Ontology-development environments Protégé 7/5/2004 ICWE2004, Munich, Germany 33 What lack ontology building tools? Shortcomings Ontologies built on AI concepts Tools and languages don t use the same terminology Software practitioners don t know all these ontology concepts They need more familiar notation and tools They need a unified representation for ontologies UML as a natural solution 7/5/2004 ICWE2004, Munich, Germany 34 17

22 Model Driven Architecture: An Overview Model Driven Architecture Developing in parallel with Semantic Web Object Modeling Group effort The latest paradigm shift in software engineering (Bézivin, 2002) From OO technology To model technology Model Driven Development (Mellor et al, 2003) 7/5/2004 ICWE2004, Munich, Germany 2 1

23 What is a model? OMG is about models Model - a simplified view of reality, or more formally A model is a set of statements about some system under study (Seidewitz, 2003) 7/5/2004 ICWE2004, Munich, Germany 3 Why do we need models? We express models using concepts that are much less bound to the underlying implementation technology and are much closer to the problem domain relative to most popular programming languages. (Selic, 2003) 7/5/2004 ICWE2004, Munich, Germany 4 2

24 Why do we need models? J2EE.NET Use of platform independent models (PIM) as specification Transformation into platform specific models (PSM) using tools 7/5/2004 ICWE2004, Munich, Germany 5 Model characteristics To be useful and effective model must possess the following characteristics: abstraction understandability accuracy predictiveness inexpensive 7/5/2004 ICWE2004, Munich, Germany 6 3

25 What is a metamodel? A metamodel makes statements about what can be expressed in the valid models of a certain modeling language. (Seidewitz, 2003) 7/5/2004 ICWE2004, Munich, Germany 7 What is a metamodel? Metamodel defines the correspondence between a model and a system In fact, A metamodel is a model of a modeling language, or A metamodel makes statements about what can be expressed in the valid models of a certain modeling language System represents Model Metamodel terminology assertions 7/5/2004 ICWE2004, Munich, Germany 8 4

26 Model Driven Architecture Common understanding of the four-layer architecture 7/5/2004 ICWE2004, Munich, Germany 9 Metametamodel Meta-Object Facility (MOF) MDA metametamodel layer (M3) Self-defined language Standard metamodeling framework for model and metadata driven systems, e.g., Modeling and development tools Data warehouse systems Metadata repositories Metadata = data about data, e.g., database schemas 7/5/2004 ICWE2004, Munich, Germany 10 5

27 Metametamodel Meta-Object Facility (MOF) MOF is the MDA s basic mechanism for defining modeling languages at the metamodel layer (M2): Unified Modeling Language (UML) Common Warehouse Metamodel (CWM) Ontology Definition Metamodel (ODM) MOF uses the UML graphical notation 7/5/2004 ICWE2004, Munich, Germany 11 UML Metamodel A part of UML metamodel 7/5/2004 ICWE2004, Munich, Germany 12 6

28 Element ModelElement +constrainedelement name : Name n {ordered} ElementOwnership n +ownedelement visibility : VisibilityKind isspecification : Boolean +constraint +namespace 0..1 n Feature Namespace GeneralizableElement Parameter Constraint ownerscope : ScopeKind isroot : Boolean defaultvalue : Expression body : BooleanExpression visibility : VisibilityKind isleaf : Boolean kind : ParameterDirectionKind isabstract : Boolean n +feature n +typedparameter n +parameter +owner 0..1 Classifier 1 +type {ordered} 1 +type StructuralFeature BehavioralFeature 0... {ordered} multiplicity : Multiplicity changeability : ChangeableKind n +typedfeature isquery : Boolean targetscope : ScopeKind ordering : OrderingKind Attribute initialvalue : Expression Operation concurrency : CallConcurrencyKind 1 n Method body : ProcedureExpressio... isroot : Boolean +specification isleaf : Boolean isabstract : Boolean specification : String 7/5/2004 ICWE2004, Munich, Germany 13 Models Model layer (M1) the layer where we develop real-world models (or domain models). in terms of UML models, that means creating classes, their relations, states, etc. 7/5/2004 ICWE2004, Munich, Germany 14 7

29 The bottom-most layer What about the bottom-most layer in MDA? There are two approaches: 1. The bottom layer is the instance layer (M0) (common understanding) 2. The instance layer contains things from our reality concrete and abstract (Atkinson & Kühne, 2003) 7/5/2004 ICWE2004, Munich, Germany 15 Orthogonal instantiation relations More details about the second approach UML classes and and objects are instances of UML metamodel concepts objects model concrete real-world things (e.g. Mike) classes model abstract real-world (e.g. Person) however, an object (Mike) is an instances of a class (Person)! 7/5/2004 ICWE2004, Munich, Germany 16 8

30 Orthogonal instantiation relations MDA orthogonal instantiation axis: linguistic instantiation (e.g. between MOF and UML metamodel) ontological instantiation (e.g. between a UML class and a UML object) Linguistic instantiation M3 M2 M1 M0 Metametamodel (MOF) Metamodel (UML, CWM) Models, model instances Real-world things X M I Ontological instantiation 7/5/2004 ICWE2004, Munich, Germany 17 Orthogonal instantiation relations Ontological instantiation one linguistic layer (e.g. M1) may have many ontological layer example: Petri net ontology UML metamodel M2 Module ModuleInstance concrete module instances M1 O2 O1 O0 Ontological instantiation M0 7/5/2004 ICWE2004, Munich, Germany 18 9

31 Why we need UML Profiles? UML Profiles we can develop a new domain language (i.e. the metamodel) using MOF it is rather expensive to develop new tools so, we try to use existing, well-known tools most of MDA tools are oriented towards UML 7/5/2004 ICWE2004, Munich, Germany 19 Why we need UML Profiles? UML Profiles a way to adapt UML to specific domains extend the UML metamodel UML as a family of languages Some UML Profiles UML Profile for CORBA UML Profile for Enterprise Application Integration UML Profile for DTD and XML Schema definition 7/5/2004 ICWE2004, Munich, Germany 20 10

32 UML Profiles Extension mechanism stereotypes, tagged values, andocl constraints UML2 improved support for profiles 7/5/2004 ICWE2004, Munich, Germany 21 MDA metadata sharing XML Metadata Interchange (XMI) a standard way of mapping objects to XML XML is not object-oriented uses MOF as the underlying object model defines two sets of rules 1. for serializing objects to XMI documents 2. for generating XML Schemas from models 7/5/2004 ICWE2004, Munich, Germany 22 11

33 MDA metadata sharing XML Metadata Interchange (XMI) may be used for serializing objects at different meta-levels (e.g., data, metadata, metametadata, etc.) important: XMI is not just for UML a UML tool will usually only accept UML XMI (i.e., XMI conforming to the UML metamodel) 7/5/2004 ICWE2004, Munich, Germany 23 XML Metadata Interchange (XMI) Mapping MDA metametamodel, metamodels, and models to XMI New metamodels (e.g. for ontologies) New domain models 7/5/2004 ICWE2004, Munich, Germany 24 12

34 XML Metadata Interchange (XMI) XMI Schema production rules we didn t show XMI production rules from MOF-compliant (meta)models see (Grose et al, 2002) for details some important documents from the OMG site XML schema for MOF metamodels XMI document of the MOF metamodel XMI document of the UML metamodel XML schema for UML models benefits and problems 7/5/2004 ICWE2004, Munich, Germany 25 13

35 UML-based solutions and tools for ontology development UML & Ontologies Ontologies frame-based systems, description logics a well-understood semantic basis the main component of the Semantic Web XML/RDF-based languages (e.g. OWL, DAML+OIL, OIL, ) concepts are unknown to practitioners lack of graphical notation 7/5/2004 ICWE2004, Munich, Germany 2 1

36 UML & Ontologies UML object oriented modeling paradigm widely accepted modeling language well-known graphical notation Object Constraint Language (OCL) many software tools industry and research XML for sharing models (XMI) 7/5/2004 ICWE2004, Munich, Germany 3 The first approach for using UML in ontology development Cranefield firstly addressed the problem the starting points: UML class diagrams provide a static modeling capability that is well-suited for representing ontologies UML object diagrams can be interpreted as declarative representations of knowledge OCL for ontology constraints advantage to use the same paradigm for modeling ontologies and knowledge 7/5/2004 ICWE2004, Munich, Germany 4 2

37 The Cranefield s approach Shortcomings of UML it lacks a formal definition defined by a metamodel some additional constraints expressed in a semi formal language (OCL) descriptions of the various elements of the language in English there are initiatives that try to solve this problem (e.g Precise UML - PUML) 7/5/2004 ICWE2004, Munich, Germany 5 The Cranefield s approach Technology requirements XMI for sharing UML models RDF for sharing ontologies UML tools that produce UML XMI XSLT that transform UML XMI to: a set of Java classes and interfaces corresponding to those in the ontology RDF & RDF Schema 7/5/2004 ICWE2004, Munich, Germany 6 3

38 The Cranefield s approach Example: The family ontology 7/5/2004 ICWE2004, Munich, Germany 7 The Cranefield s approach Example: transformation to RDF(S) XSLT implementation Classes to RDFS Objects to RDF Mapping problems: UML classes have different features attributes, associations, and association classes RDFS fields or properties RDFS properties are first-class objects 7/5/2004 ICWE2004, Munich, Germany 8 4

39 The Cranefield s approach Transformation to RDFS Some solutions: properties in RDFS have a class prefix (but, this has a problem with class inheritance) a multiplicity upper limit greater than RDFS bag association ends with a UML ordered constraint RDFS sequences 7/5/2004 ICWE2004, Munich, Germany 9 The Cranefield s approach Resulting RDFS for the family ontology (excerpt) <?xml version="1.0" encoding="utf-8"?> <rdf:rdf xml:lang="en" xmlns:rdfsx=" xmlns:rdfs=" xmlns:rdf=" <rdfs:class rdf:id=" <rdf:property ID=" <rdfs:domain rdf:resource=" <rdfs:range rdf:resource="rdfs:literal"/> </rdf:property> <rdf:property ID=" <rdfs:domain rdf:resource=" <rdfs:range rdf:resource="rdf:bag"/> <rdfsx:containerelementtype rdf:resource=" </rdf:property> <rdf:property ID=" <rdfs:domain rdf:resource=" <rdfs:range rdf:resource="rdf:seq"/> <rdfsx:containerelementtype rdf:resource=" </rdf:property> <rdf:property ID=" <rdfs:domain rdf:resource=" <rdfs:range rdf:resource=" </rdf:property> 7/5/2004 ICWE2004, Munich, Germany 10 5

40 The Cranefield s approach Some limitation of this approach diagrams are just notation feature standard UML can not express advanced ontology features (e.g. restrictions) one cannot conclude whether the same property was attached to more than one class 7/5/2004 ICWE2004, Munich, Germany 11 The Cranefield s approach Some limitation of this approach one cannot create a hierarchy of properties target RDFS ontology description does not have advanced restriction concepts (e.g. multiplicity) However, this is a kind of agile UML-based ontology development 7/5/2004 ICWE2004, Munich, Germany 12 6

41 UML extension for ontology development (Backlawski et al, 2002) Rough classification of KR languages: Logical languages (e.g. KIF) Frame-based languages OO database languages Graph-based languages conceptual graphs (Sowa, 2000) This work is inspired by the DAML+OIL Theoretical discussion 7/5/2004 ICWE2004, Munich, Germany 13 UML extension for ontology development DAML+OIL vs. UML DAML+OIL don t have any graphical notation For DAML+OIL there are no many tools or experiences as there is with UML UML and DAML+OIL have many characteristic in common (e.g classes and instances) Property as a stand alone concept the main problem Solution that use stereotypes do not define semantic (e.g. DUET) 7/5/2004 ICWE2004, Munich, Germany 14 7

42 UML extension for ontology development DAML+OIL and UML: Some differences Specialization/generalization UML behavioral DAML+OIL set-theoretic e.g. Square and Rectangle not semantically equivalent, but mapping is consistent UML is not monotonic while DAML+OIL is monotonic KR languages do not have meta-level separation (e.g. instances of a class may be class) 7/5/2004 ICWE2004, Munich, Germany 15 UML extension for ontology development DAML+OIL vs. UML High-level mapping between DAML+OIL and UML DAML Concept Ontology Class As Sets (union, intersection, etc.) Hierarchy Property Hierarchy Restriction Data Types Instances and Values Similar UML Concept Package Class Not supported Class Generalization Relations Aspects of Attributes, Associations ans Classes None for Attributes, lim ited Generalization for Associations, Class Generalization for Relations Constraints Association Ends, including m ultiplicity and roles, Im plicitly as class containing the attribute Data Types Object Instances and Attribute Values 7/5/2004 ICWE2004, Munich, Germany 16 8

43 UML extension for ontology development Extension of the UML metamodel GeneralizableElement Classifier Association Property DerivedClassifier Class DifferentIndividualFrom EquivalentTo Union Complement SameClassAs SameIndividualAs Int ersec tion Restriction SamePropertyAs 7/5/2004 ICWE2004, Munich, Germany 17 UML extension for ontology development Extension of the UML metamodel Association 1 2..* AssociationEnd 0..* 0..* 0..1 Property transitive : boolean 1..* +participant 1 Classifier complementof intersectionof 0..* 0..* toclass 1..* 0..* unionof onproperty 0..* Restriction 0..* Complement Intersection Union 7/5/2004 ICWE2004, Munich, Germany 18 9

44 UML extension for ontology development Example Faculty +affiliation:location Organization <<Property>> location onproperty <<Restriction>> toclass <<Union>> <<Intersection>> University School Institute <<Complement>> Company 7/5/2004 ICWE2004, Munich, Germany 19 UML extension for ontology development Some shortcomings DAML+OIL is evolved to OWL tool support is not given extended UML metamodel is an awkward solution we do not need all UML concepts for ontology engineering Better solution an independent ontology metamodel 7/5/2004 ICWE2004, Munich, Germany 20 10

45 Unified Ontology Language (UOL) A new MOF-defined ontology metamodel (Backlawski et al, 2002) closely related to UML direct mappings with UML based on UML and DAML+OIL (but, also considers OWL) UOL uses the same analysis about KR languages 7/5/2004 ICWE2004, Munich, Germany 21 Unified Ontology Language (UOL) The proposed language UOL should satisfy the following requirements: it must be a MOF metalanguage a bounded two-way mapping between core UML and core UOL. The two-way mapping must preserve semantic equivalence on levels 0 and 1 of MDA 7/5/2004 ICWE2004, Munich, Germany 22 11

46 Unified Ontology Language (UOL) The proposed language UOL should satisfy the following requirements: Core UML and core UOL must include the following notions: Package Class Binary association Generalization Attribute Multiplicity constraints 7/5/2004 ICWE2004, Munich, Germany 23 Unified Ontology Language (UOL) MOF-compliant metamodel of UOL Node object 1 Fact 0..n 0..n subject 1 0..n predicate 0..n 1 Resource 0..n Literal context 1 1 context Ontology Package 0..n imports 0..n 7/5/2004 ICWE2004, Munich, Germany 24 12

47 Unified Ontology Language (UOL) Short discussion a small and high-level metamodel a solid starting point for a standard MOF-based ontology language we have been unable to find practical software tool that support this proposal and mapping to the Semantic Web languages 7/5/2004 ICWE2004, Munich, Germany 25 Transformation-Based Approach (Falkovych et al, 2003) transforming domain models between UML and semantic web languages focusing on the actual transformation between the two languages 7/5/2004 ICWE2004, Munich, Germany 26 13

48 Transformation-Based Approach Starting presumptions UML and ontology languages complement one another UML is designed for model building by human experts OWL is designed to be used at run time to provide guidance for intelligent processing methods the translation is less than trivial, because of the differences between the two languages 7/5/2004 ICWE2004, Munich, Germany 27 Transformation-Based Approach Extraction of ontologies from UML diagrams transformation of UML diagrams into DAML+OIL ontologies with preserving the semantics of UML concepts that there exist large sources of ontological knowledge already available in design documents of existing applications reasoning about these (UML-based) ontologies allow knowledge interchange between heterogeneous environments 7/5/2004 ICWE2004, Munich, Germany 28 14

49 Transformation-Based Approach Property mappings the work is based on the UML extension proposed by Backlawski et al especially on the Property problem an association it is mapped into daml:objectproperty an attribute is mapped into daml:datatypeproperty 7/5/2004 ICWE2004, Munich, Germany 29 Transformation-Based Approach Taxonomy of association types to preserve semantics of UML associations to distinguish association ends Association Binary Unidirectional Aggregation Composition Following_direction Whole_of_part C_whole_of_part Opposite_to_direction Part_of_whole C_part_of_whole 7/5/2004 ICWE2004, Munich, Germany 30 15

50 Transformation-Based Approach Taxonomy of association types all association types can have a name or be unnamed they can have role names attached to association ends All associations in a particular diagram are modeled as sub-properties of corresponding association types from the taxonomy 7/5/2004 ICWE2004, Munich, Germany 31 Transformation-Based Approach Example: Unnamed <daml:class binary rdf:id="boundary association point"> Boundary point 1 +has 1..n +form Boundary segment <rdfs:label>boundary point</rdfs:label> <rdfs:subclassof> <daml:restriction daml:mincardinality="1"> <daml:onproperty rdf:resource="#has_g.4"/> <daml:toclass rdf:resource="#boundary segment"/> </daml:restriction> </rdfs:subclassof> <rdfs:subclassof> <daml:restriction daml:mincardinality="1"> <daml:onproperty rdf:resource="#_g.2"/> <daml:toclass rdf:resource="#boundary segment"/> </daml:restriction> </rdfs:subclassof> </daml:class> <daml:objectproperty rdf:id="has_g.4"> <rdfs:subpropertyof rdf:resource="#_g.2"/> </daml:objectproperty> <daml:objectproperty rdf:id="_g.2"> <rdfs:subpropertyof rdf:resource="#binary_unnamed"/> </daml:objectproperty> 7/5/2004 ICWE2004, Munich, Germany 32 16

51 Transformation-Based Approach Example: Unnamed <daml:class binary rdf:id="boundary association segment"> Boundary point 1 +has 1..n +form Boundary segment <rdfs:label>boundary segment</rdfs:label> <rdfs:subclassof> <daml:restriction daml:cardinality="1"> <daml:onproperty rdf:resource="#form_g.3"/> <daml:toclass rdf:resource="#boundary point"/> </daml:restriction> </rdfs:subclassof> <rdfs:subclassof> <daml:restriction daml:cardinality="1"> <daml:onproperty rdf:resource="#g.2"/> <daml:toclass rdf:resource="#boundary point"/> </daml:restriction> </rdfs:subclassof> </daml:class> <daml:objectproperty rdf:id="form_g.3"> <rdfs:subpropertyof rdf:resource="#g.2"/> </daml:objectproperty> <daml:objectproperty rdf:id="g.2"> <rdfs:subpropertyof rdf:resource="#binary_unnamed"/> </daml:objectproperty> 7/5/2004 ICWE2004, Munich, Germany 33 Transformation-Based Approach Characteristics the approach preserves the semantics of UML association in DAML+OIL one can check consistency of UML models using DL-based reasoners (e.g. FaCT) deriving new knowledge by using inference engines XSLT implementation (from XMI to DAML+OIL) some experiences: XSLT is quite cumbersome for more complex mappings very sensitive to the format of the input file 7/5/2004 ICWE2004, Munich, Germany 34 17

52 Protégé Protégé and software engineering a leading ontology editor a number of different plug-in XML, DAML+OIL (backends), and OIL (tab) integrate ontology design into mainstream software projects (Knublauch, 2003) Protégé objects can be used as Java objects ontology languages have extended declarative power that can improve object oriented models 7/5/2004 ICWE2004, Munich, Germany 35 Protégé and software engineering Protégé s metamodel MOF-based metamodel on the same metalevel as UML Protégé basically implements an adapted version of the OKBC metamodel improved interoperability with other systems extensible and adaptable e.g. a new ontology language can bee added 7/5/2004 ICWE2004, Munich, Germany 36 18

53 Protégé s MOF-based metamodel 7/5/2004 ICWE2004, Munich, Germany 37 Protégé and software engineering Protégé s MOF-based plug-ins XMI backend the XMI format of the Protégé s metamodel UML backend the UML XMI format these two backends use the NetBeans MetaData Repository (MDR) 7/5/2004 ICWE2004, Munich, Germany 38 19

54 Protégé and software engineering Protégé s MOF-based plug-ins MOF models can be accessed in Java using Java Metadata Interface (JMI) JMI takes a MOF metamodel and generates Java interfaces this can be an alternative way for creating a Protégé knowledge base 7/5/2004 ICWE2004, Munich, Germany 39 Protégé s UML backend 7/5/2004 ICWE2004, Munich, Germany 40 20

55 Protégé and software engineering Protégé s UML backend Protégé to UML mapping Classes Classes Inheritance Inheritance Primitive Slots Attributes Instance Slots Associations UML to Protégé mapping the name of a new slot is already taken by a different slot: <attributename>@<classname> 7/5/2004 ICWE2004, Munich, Germany 41 Protégé and software engineering Protégé s UML backend UML extensions (UML Profiles) are supported only unidirectional (for export) UML tools Poisedon for UML v1.6 (recommended) Rational Rose MagicDraw 6.0 Together Control Center 6.0 7/5/2004 ICWE2004, Munich, Germany 42 21

56 DAML UML Enhanced Tool (DUET) DUET characteristics UML visualization and authoring environment for DAML core DAML concepts are being mapped into UML through a UML profile for DAML capability to work with multiple ontologies simultaneously Implementation: an add-in to Rational Rose for use with ArgoUML 7/5/2004 ICWE2004, Munich, Germany 43 DUET Rational Rose Add-in 7/5/2004 ICWE2004, Munich, Germany 44 22

57 DAML UML Enhanced Tool (DUET) From UML to DAML UML <<DAMLontology>> Package Class <<DAMLproperty>> Class <<DAMLdatatype>> Class Data Type Associations and Roles Attributes Generalization Import Fully Qualified Names DAML Ontology Class Property Datatype XML Schema or DAML datatype Object Property and Restriction Datatype Property subclassof or subpropertyof Import Namespaces 7/5/2004 ICWE2004, Munich, Germany 45 Xpetal Transformation of UML models to RDF(S) Java implementation similar mapping rules as the Cranefield s XSLT only for Rational Rose models in the mdl format it is not based on standards (i.e. XMI) 7/5/2004 ICWE2004, Munich, Germany 46 23

58 Short summary of analyzed approaches Approach Metamodel Model description Transformation mechanism Generated ontology language Cranefield Standard UML UML XMI XSLT RDFS, Java classes Backlawski et al UML Profile, MOF-based ontology language not given (XMI) - DAML Falkovych at al Standard UML UML XMI XSLT DAML + OIL Protégé DUET Xpetal Protégé metamodel Standard UML UML Profile Standard UML Protégé XMI UML XMI Rational Rose, ArgoUML Rational Rose mdl files Programmed Programmed Programmed OWL, RDF(S), DAML+OIL, XML, UML XMI, Protégé XMI, DAML+OIL RDFS 7/5/2004 ICWE2004, Munich, Germany 47 UML-based ontology development Short summary all these approaches are useful none of them contains: a full description of the new MDA-based ontology language a related UML Profile transformations between these two languages transformations to Semantic Web languages solution: the OMG s initiative for Ontology Definition Metamodel (ODM) 7/5/2004 ICWE2004, Munich, Germany 48 24

59 OMG s Initiative: Ontology Definition Metamodel and Current Submissions OMG s Request for Proposal (RFP) UML could be a means towards more rapid development of ontologies: the familiarity of users with UML the availability of UML tools the existence of many domain models in UML the similarity of those models to ontologies a number of tools and investigations have demonstrated the practicality of using UML based tools for developing ontologies 7/5/2004 ICWE2004, Munich, Germany 2 1

60 OMG s Request for Proposal (RFP) This approach continues the Object Management Group's "gradual move to more complete semantic models". It would also create a link between the UML community and the emerging Semantic Web community, much as other metamodels and profiles have created links with the developer and middleware communities. OMG Document: ad/ /5/2004 ICWE2004, Munich, Germany 3 OMG s RFP: Ontology Definition Metamodel (ODM) Several specifications should be defined : a standard MOF2 compliant metamodel for Ontology Definition Metamodel (ODM) PIM a UML2 Profile to support reuse of UML notation for ontology definition a mapping from the Ontology Definition Metamodel to the profile, and a language mapping for the ODM to the W3C Web Ontology Language for Description Logics, OWL DL 7/5/2004 ICWE2004, Munich, Germany 4 2

61 OMG s RFP: Ontology Definition Metamodel (ODM) Graphical representation of RFP PIM OWL PSM 7/5/2004 ICWE2004, Munich, Germany 5 ODM Specification Requirements Mandatory Requirements define an Ontology Definition Metamodel (ODM) using MOF2 Core that represents the semantics of ontologies including but not necessarily limited to OWL ontologies depict ODM using UML it must be based on the appropriate elements of 2.x versions of OMG metamodels including, MOF, UML, and OCL. The resulting metamodel shall be MOF2-compliant a UML2 Profile extending the UML2 metamodel for ontology definition 7/5/2004 ICWE2004, Munich, Germany 6 3

62 ODM Specification Requirements Mandatory Requirements forward and reverse engineering of logically equivalent ontologies between environments to support iterative development of ontologies a language mapping for ODM to the W3C OWL DL language This mapping shall be two-way and bounded an XMI for MOF2.0 compliant W3C XML Schema Description (xsd) based on ODM 7/5/2004 ICWE2004, Munich, Germany 7 ODM Specification Requirements Optional Requirements mapping multiple ontologies into a single UML model round-trip engineering across language environments a mapping for DAML+OIL, or other ontology languages, with the same characteristics as that described for OWL 7/5/2004 ICWE2004, Munich, Germany 8 4

63 ODM Specification Requirements Optional Requirements a profile, metamodel, and mappings needed to support OWL Full RDF and RDFS would also be supported as a consequence design ODM to provide for specification of ontology metadata to describe context and scope of the development and intended use of a corresponding ontology 7/5/2004 ICWE2004, Munich, Germany 9 ODM Initial Submissions Submitters International Business Machines (IBM) Distributed Systems Technology Centre (DSTC) Gentleware & AT&T (sponsor) Sandpiper Software, Inc. & Knowledge Systems Laboratory (KSL), Stanford University 7/5/2004 ICWE2004, Munich, Germany 10 5

64 IBM s ODM Proposal IBM s proposal characteristics written in the well-known OMG s style Daniel Chang CWM the same terminology, constructs and semantics used in OWL DL are used in ODM satisfies almost all mandatory requirements?! the ODM XMI format is not given Mixing of meta levels is unresolved Resolution: add an OWL package for M1 use Tool support is not given Rational Rose 7/5/2004 ICWE2004, Munich, Germany 11 IBM s ODM Proposal ODM specification is grouped in packages: Kernel package Ontology package Classes package Properties package Individuals package Datatypes package 7/5/2004 ICWE2004, Munich, Germany 12 6

65 IBM s ODM Proposal Kernel package the root of the hierarchy Element +ownedelement +ontology Ontology (from Ontology) 0..n 0..1 Named Ele ment namespace : URIreference name : String / id : URIreference 7/5/2004 ICWE2004, Munich, Germany 13 IBM s ODM Proposal Classes Package Class Descriptions NamedElement (from Kernel) complementof 0..1 Thing (from Individuals) 0..n oneof Class complete : Boolean intersectionof 0..n Class unionof 0..n Restriction 7/5/2004 ICWE2004, Munich, Germany 14 7

66 IBM s ODM Proposal Classes Package Class Axioms equivalentclass 0..n Class 0..n 0..n subclassof disjointwith DeprecatedClass 7/5/2004 ICWE2004, Munich, Germany 15 IBM s ODM Proposal NamedElement (from Kernel) Properties Package subpropertyof 0..n Class (from Classes) 0..n domain Property 0..n range 0..n equivalentproperty 0..1 ObjectProperty DataRange (from Datatypes) 0..n range DatatypeProperty 0..1 inverseof SymmetricProperty TransitiveProperty InverseFunctional Property FunctionalObject Property DeprecatedObject Property FunctionalDatatype Property 7/5/2004 ICWE2004, Munich, Germany 16 DeprecatedDatatype Property 8

67 IBM s ODM Proposal Individuals package NamedElement (from Kernel) Thing differentfrom <objectpropertyid> 0..n Thing <datatypepropertyid> Literal (from Datatypes) 0..n 0..n sameas type 0..n Class (fro m Cl asses) sameindividualas 7/5/2004 ICWE2004, Munich, Germany 17 IBM s ODM Proposal Example: The Wine ontology <<Class>> Wine +hasmaker <<Class>> Winery <<Ontology>> OWL <<imports>> <<imports>> <<FunctionalProperty>> hasmaker <<Ontology>> Wine <<imports>> <<Ontology>> Food <<inverseof>> <<imports>> <<ObjectProperty>> produceswine 7/5/2004 ICWE2004, Munich, Germany 18 9

68 IBM s ODM Proposal The Wine ontology <<unionof>> <<Class>> WineDescriptor <<unionof>> <<Class>> WineTaste <<Class>> WineColor <<oneof>> <<oneof>> <<oneof>> <<Thing>> Red <<Thing>> Rose <<Thing>> White 7/5/2004 ICWE2004, Munich, Germany 19 IBM s ODM Proposal The Wine ontology <<differentfrom>> <<Thing>> Dry <<differentfrom>> <<type>> <<Thing>> OffDry <<type>> <<Class>> WineSugar <<differentfrom>> <<Thing>> Sweet <<type>> 7/5/2004 ICWE2004, Munich, Germany 20 10

69 IBM s ODM Proposal The Wine ontology <<Class>> Wine <<subclassof>> <<Restriction>> (from OWL) <<onproperty>> <<cardinality>> <<s omevaluesfrom>> <<TransitiveProperty>> locatedin <<NonNegativeInteger>> 1 <<Class>> Region 7/5/2004 ICWE2004, Munich, Germany 21 DSTC s ODM Proposal DSTC s proposal characteristics based on a very deep analysis of ontology concepts does not support some specific ontology languages OWL Full is supported Starter ontology provides a basic core of concepts a basis for other ontologies 7/5/2004 ICWE2004, Munich, Germany 22 11

70 DSTC s ODM Proposal DSTC s proposal characteristics ODM is defined as a UML extension UML Profile is not given UML profiling is not sufficient for the expressive power needed for ODM and a full MOF meta-model is presented a full MOF metamodel enables the use MDA tools repository generators XMI generators Query/View/Transformation tools 7/5/2004 ICWE2004, Munich, Germany 23 DSTC s ODM Proposal Packages Kernel Ontology Root Representations Multiplicities Expressions Statements Classifiers Classes 7/5/2004 ICWE2004, Munich, Germany 24 12

71 DSTC s ODM Proposal The Root package Ontology::Root +created by 1 Kernel::Element Kernel::Comment ElementWithCreator TaggedComment 7/5/2004 ICWE2004, Munich, Germany 25 DSTC s ODM Proposal The Root package Ontology::Representations Kernel::AggregationKind Kernel::Property AggregationKind member substance Kernel::Class Property isrigid : Boolean [0..1] IdentityClass Representation Dimension sameas( InstanceSpec) : boolean 7/5/2004 ICWE2004, Munich, Germany 26 13

72 DSTC s ODM Proposal The developed prototypes to validate parts of the specification Web-KB a non-mof-based implementation of many of the concepts represented in this proposal DSTC s dmof product (MOF 1.3) TokTok product (HUTN 1.0) 7/5/2004 ICWE2004, Munich, Germany 27 DSTC s ODM Proposal Short summary the proposed metamodel is a UML extension UML Profile is not defined ODM XMI is not defined it will be defined in the final version metamodel mapping rules to produce XMI two-way mapping is not defined the submitters think that this is not a problem mapping between ODM and OWL MOF2 Query/View/Transformation 7/5/2004 ICWE2004, Munich, Germany 28 14

73 Gentleware s ODM Proposal Gentleware s proposal characteristics Ontology Definition Metamodel a MOF 2.0 metamodel specializes components from the Core package of the Infrastructure Library makes reference to elements from the UML kernel UML Profile defines a small set of stereotypes and tag-values for the modeling of OWL ontologies ODM XMI is not provided 7/5/2004 ICWE2004, Munich, Germany 29 Gentleware s ODM Proposal Gentleware s proposal characteristics Proposal contains two level metamode: language independent metamodel!? OWL-dependent metamodel DUET tool prototype software support Rational Rose add-in ArgoUML plug-in it is not a full support for Gentleware s UML Profiles tested on well-known several ontologies (e.g.wordnet) RDF 7/5/2004 ICWE2004, Munich, Germany 30 15

74 Gentleware s ODM Proposal Package organization <<metamodel>> MOF <<instanceof>> <<ins tanceof>> <<metamodel>> UML <<metamodel>> ODM (from Ontologies) <<extends>> <<profile>> Profile (from Ontologies) <<metamodel>> OWL (from Ontologies) 7/5/2004 ICWE2004, Munich, Germany 31 Gentleware s ODM Proposal Resource RDF(S)-based approach +extent 0..n +type 1 * +general Classifier 1 (from Constructs) type <<metaclass>> Resource <<1>> URI : String TypedElement (from Constructs) 1 1 +object 1 +subject +import 0..n <<metaclass>> Ontology <<metaclass>> DataType <<metaclass>> Class 0..n +subpropertyof 0..n <<metaclass>> Property <<metaclass>> Literal +subclassof +/properties 0..n 0..n <<metaclass>> Instance 7/5/2004 ICWE2004, value : String Munich, Germany n 16

75 Gentleware s ODM Proposal OWL extensions to ODM Core package <<metaclass>> Class (from ODM) <<metaclass>> Property (from ODM) <<metaclass>> SymetricProperty <<metaclass>> InverseFunctionalProperty <<metaclass>> Restriction +onproperty 1 <<metaclass>> Property (from ODM) <<metaclass>> TransitiveProperty <<metaclass>> FunctionalProperty 7/5/2004 ICWE2004, Munich, Germany 33 <<ontology>> Wine Gentleware s ODM Proposal The Wine ontology (example) Wine 0..n 1 Winery <<imports>> hasmaker <<instanceof>> <<property>> FunctionalProperty (from OWL) <<ontology>> OWL <<instanceof>> <<property>> inverseof (from OWL) <<references>> <<property>> produces Wine <<ontology>> RDFS 7/5/2004 ICWE2004, Munich, Germany 34 17

76 Gentleware s ODM Proposal The Wine ontology <<collection>> _WineDescriptor WineTaste : WineDescriptor WineColor : WineDescriptor <<unionof>> <<Class>> WineDescriptor <<unionof>> IBM 0..* unionof WineDescriptor <<Class>> WineTaste <<Class>> WineColor WineTaste WineColor <<oneof>> <<oneof>> oneof 0..* <<oneof>> _WineColor White : WineColor Red : WineColor Rose : WineColor <<Thing>> Red <<Thing>> Rose 7/5/2004 ICWE2004, Munich, Germany 35 <<Thing>> White Gentleware s ODM Proposal The Wine ontology <<Class>> Wine _:Restriction _:onproperty _:locatedin <<subclassof>> _:somevaluesfrom _:cardinality <<Restriction>> (from OWL) _:Region _:xsd:nonnegativeinteger value = '1' <<onproperty>> <<cardinality>> <<s omevaluesfrom>> <<Transiti vepropert y>> locatedin <<NonNegativeInteger>> 1 <<Class>> Region 7/5/2004 ICWE2004, Munich, Germany 36 18

77 Sandpiper Software and KSL s ODM Proposal Sandpiper Software and KSL s proposal characteristics deep AI technology basis (KIF), KL-ONE, Ontolingua, Chimaera, LOOM, Generic Frame Protocol (GFP) Visual Ontology Modeler (VOM) Rational Rose add-in ontology-language independent ODM 7/5/2004 ICWE2004, Munich, Germany 37 Sandpiper Software and KSL s ODM Proposal Sandpiper Software and KSL s proposal characteristics Open Knowledge Base Connectivity as a basis for ODM an interchange language for frame-based knowledge bases: LOOM (ISI), Ontolingua/KIF, DAML+OIL, OWL, the Protégé s metamodel the FIPA standard metamodel for ontology services 7/5/2004 ICWE2004, Munich, Germany 38 19

78 Sandpiper Software and KSL s ODM Proposal Layered language specification DAML-S OWL Full DAML+OIL PSL SWSL RDF Schema Metamodel CL / KIF Metamodel RDF Metamodel OKBC Extended Metamodel and Services OKBC Core Metamodel 7/5/2004 ICWE2004, Munich, Germany 39 Sandpiper Software and KSL s ODM Proposal <<ontology Class>> entity <<ontologyclass>> arrangement OKBC Core Metamodel <<ontologyclass>> boolean <<ontologyclass >> collection <<ontologyclass>> frame <<ontologyclass>> symbol <<ontologyclass>> number <<ontologyclass>> bag <<ontologyclass>> set <<ontologyclass>> list <<ontologyclass>> realnumber <<ontologyclass>> complexnumber <<ontologyclass>> relation <<ontologyclass>> string <<ontologyclass>> integer 7/5/2004 ICWE2004, Munich, Germany 40 20

79 Sandpiper Software and KSL s ODM Proposal OKBC Core Metamodel <<ontologyclass>> relation <<ontologyclass>> unaryrelation <<ontologyclass>> binaryrelation <<ontologyclass>> function <<ontologyclass>> ternaryrelation <<ontologyclass>> variablearityrelation <<ontologyclass>> individual <<ontologyclass>> class <<ontologyclass>> slotrelation <<ontologyclass>> facet 7/5/2004 ICWE2004, Munich, Germany 41 Sandpiper Software and KSL s ODM Proposal Short summary The foundation ontology The minimal number of ODM construct necessary for ontology development ODM XMI is not given mappings between ODM and OUP, and ODM and OWL are not given 7/5/2004 ICWE2004, Munich, Germany 42 21

80 Summary ODM Proposals Metamodels Proposal Ontology languageindependent MOFcompliant metamodel UML Extension Explicit UML Ontology Profile OWLdependent metamodel IBM DSTC Gentleware / /+ Sandpiper & KSL + + 7/5/2004 ICWE2004, Munich, Germany 43 Summary ODM Proposals Mappings, ODM XMI and tool support Proposal Similarity with other OMG s specifications ODM XMI is given ODM OUP OUP ODM ODM OWL OWL ODM Tool support IBM Not stated DSTC -/ WebKB2 Gentleware -/+ - -/+ -/+ -/+ -/+ DUET Sandpiper & KSL VOM 7/5/2004 ICWE2004, Munich, Germany 44 22

81 The MDA-Based Infrastructure for Ontology Development GoodOldAI efforts The GoodOldAI research group approaching AI and software engineering different AI knowledge representation techniques frame systems, logics, rules, fuzzy systems, neural networks, Petri nets, OBOA Object-oriented AIR Model Driven Architecture MDA-based ontology development a part of these efforts 7/5/2004 ICWE2004, Munich, Germany 2 1

82 OBOA / GET-BITS OBOA - OBject-Oriented Abstraction (Devedzic et al., 1996-present) a model of intelligent systems a framework for developing intelligent systems a set of software components for building intelligent systems a layered reference architecture and ontology of intelligent systems 7/5/2004 ICWE2004, Munich, Germany 3 OBOA / GET-BITS OBOA - OBject-Oriented Abstraction (Devedzic et al., 1996-present) Level of abstraction Objective Semantics Level of abstraction Dimensions D 1 D 2... D n Level 1 Integration Multiple agents or systems Level 1 Level 2 System Single agent or system Level 2 Level 3 Blocks System building blocks Level 3 Level 4 Units Units of blocks Level 4 Level 5 Primitives Parts of units Level 5 (a) (b) 7/5/2004 ICWE2004, Munich, Germany 4 2

83 AIR Artificial Intelligence Research AIR 7/5/2004 ICWE2004, Munich, Germany 5 AIR Conceptual solution ODM 7/5/2004 ICWE2004, Munich, Germany 6 3

84 AIR Conceptual solution 7/5/2004 ICWE2004, Munich, Germany 7 MDA-Based Ontology Infrastructure 7/5/2004 ICWE2004, Munich, Germany 8 4

85 Ontology Definition Metamodel (ODM) Resources 7/5/2004 ICWE2004, Munich, Germany 9 Ontology Definition Metamodel (ODM) Decorator 7/5/2004 ICWE2004, Munich, Germany 10 5

86 Ontology Definition Metamodel (ODM) Properties Decorator 7/5/2004 ICWE2004, Munich, Germany 11 Ontology Definition Metamodel (ODM) Statements 7/5/2004 ICWE2004, Munich, Germany 12 6

87 Ontology UML Profile 7/5/2004 ICWE2004, Munich, Germany 13 7/5/2004 ICWE2004, Munich, Germany 14 7

88 Ontology UML Profile Properties 7/5/2004 ICWE2004, Munich, Germany 15 Ontology UML Profile Statements 7/5/2004 ICWE2004, Munich, Germany 16 8

89 The Wine ontology Well-known example Tagged value odm.anonymous 7/5/2004 ICWE2004, Munich, Germany 17 The Wine ontology 7/5/2004 ICWE2004, Munich, Germany 18 9

90 The Wine ontology 7/5/2004 ICWE2004, Munich, Germany 19 The Petri net ontology Modular organization (example) 7/5/2004 ICWE2004, Munich, Germany 20 10

91 The Petri net ontology Core package 7/5/2004 ICWE2004, Munich, Germany 21 The Petri net ontology Core package - enumeration 7/5/2004 ICWE2004, Munich, Germany 22 11

92 The Petri net ontology Core package - properties 7/5/2004 ICWE2004, Munich, Germany 23 The Petri net ontology Core package - restrictions 7/5/2004 ICWE2004, Munich, Germany 24 12

93 The Upgraded Petri net ontology Extension package new restrictions 7/5/2004 ICWE2004, Munich, Germany 25 The Upgraded Petri net ontology Extension package new classes and properties 7/5/2004 ICWE2004, Munich, Germany 26 13

94 Mappings Mappings: ODM<->OUP, ODM<->OWL 7/5/2004 ICWE2004, Munich, Germany 27 Mappings Important facts for mappings technological spaces (Kurtev et al, 2002) ontologies, MDA, and XML epistemological equivalence between technological layers implementation techniques XSLT, MOF2 QVT, RDQL, Java 7/5/2004 ICWE2004, Munich, Germany 28 14

95 Epistemological equivalences between technology layers 7/5/2004 ICWE2004, Munich, Germany 29 An example of a mapping A transformation in the XML technological space the transformation of OUP into OWL 7/5/2004 ICWE2004, Munich, Germany 30 15

96 An example of a mapping Transformations in the MDA technological space transformations between OUP and ODM 7/5/2004 ICWE2004, Munich, Germany 31 Transformation techniques Implementation technologies for the transformations Target language ODM OUP OWL Source language ODM XML TS MDA TS XML TS XSLT QVT XSLT OUP XML TS MDA TS XML TS XSLT QVT XSLT XML TS XML TS OWL Programmed *, Programmed *, XSLT XSLT 7/5/2004 ICWE2004, Munich, Germany 32 16

97 What is implemented? Implemented tools Ontology editor an AIR plug-in JMI-compliant repository NetBeans MDR 7/5/2004 ICWE2004, Munich, Germany 33 AIR Ontology Editor 7/5/2004 ICWE2004, Munich, Germany 34 17

98 What is implemented? Implemented mappings Ontology UML Profile to OWL XSLT UML XMI to OWL XML acts as an extension of UML tools tested on Poseidon for UML 2.0 CE sharing between Poseidon for UML and Protégé Petri net ontology and Wine ontology 7/5/2004 ICWE2004, Munich, Germany 35 OUP to OWL Mapping: XSLT 7/5/2004 ICWE2004, Munich, Germany 36 18