Web Ontology Language (OWL)

Transcription

1 Web Ontology Language (OWL) Need meaning beyond an object-oriented type system RDF (with RDFS) captures the basics, approximating an object-oriented type system OWL provides some of the rest OWL standardizes constructs to capture such subtleties of meaning OWL builds on RDF, by limiting it Limiting syntax Limiting possible interpretations OWL assigns standard semantics to new terms Munindar P. Singh (NCSU) Service-Oriented Computing Fall

2 OWL in Brief Specifies classes and properties in description logic (DL) Class operators analogous to Boolean operators and, not, and or Constraints on properties: transitive,... Restrictions: constructs unique to DL OWL 1 has Species or Dialects OWL Full OWL DL OWL Lite Munindar P. Singh (NCSU) Service-Oriented Computing Fall

3 Custom Metadata Vocabularies Metadata for services and information resources presupposes custom vocabularies Need standard semantics for the metadata to remove ambiguity despite heterogeneity <Mammal r d f : ID= Mary /> <Mammal r d f : ID= John > <h a s P a r e n t r d f : r e s o u r c e = #Mary /> </Mammal> Munindar P. Singh (NCSU) Service-Oriented Computing Fall

4 Ontologies to Define Vocabulary Semantics Example of a trivial ontology defining our vocabulary Uses simple subclasses and properties Disjointness goes beyond RDF Object properties refine RDF properties; relate two objects <o w l : C l a s s r d f : I D= Mammal > <r d f s : s u b C l a s s O f r d f : r e s o u r c e= #Animal /> <o w l : d i s j o i n t W i t h r d f : r e s o u r c e= #R e p t i l e /> </ o w l : C l a s s> <o w l : O b j e c t P r o p e r t y r d f : I D= h a s P a r e n t > <r d f s : d o m a i n r d f : r e s o u r c e= #Animal /> <r d f s : r a n g e r d f : r e s o u r c e= #Animal /> </ o w l : O b j e c t P r o p e r t y> Munindar P. Singh (NCSU) Service-Oriented Computing Fall

5 Simple Inference Find a model, if any exists Given the definition for the property hasparent and the snippet <owl : Thing r d f : ID= Fido > <h a s P a r e n t r d f : r e s o u r c e= #Rover /> </owl : Thing> we can infer that Fido is an Animal Munindar P. Singh (NCSU) Service-Oriented Computing Fall

6 OWL Entities and Relationships Illustrative as a summary: not a proper OWL diagram rdfs:subclassof rdfs:class owlequivalentclass owl:equivalentproperty owl:subpropertyof owl:class owl:datatypeproperty rdf:range owl:datarange owl:disjointwith rdf:domain owl:equivalentproperty rdf:range rdf:property rdfs:datatype owl:inverseof owl:objectproperty owl:functionalproperty owl:subpropertyof owl:inversefunctionalproperty owl:symmetricproperty owl:transitiveproperty Munindar P. Singh (NCSU) Service-Oriented Computing Fall

7 Constructing OWL Classes Explicitly <owl : C l a s s r d f : ID= Mammal > <r d f s : s u b C l a s s O f r d f : r e s o u r c e= #Animal /> <owl : d i s j o i n t W i t h r d f : r e s o u r c e= #R e p t i l e /> </owl : C l a s s > Anonymously via formal expressions using operators analogous to set operators: intersectionof, unionof, complementof <owl : C l a s s r d f : ID= SugaryBread > <owl : i n t e r s e c t i o n O f r d f : parsetype = C o l l e c t i o n > <owl : C l a s s r d f : about= #Bread /> <owl : C l a s s r d f : about= #SweetFood /> </owl : i n t e r s e c t i o n O f > </owl : C l a s s > Munindar P. Singh (NCSU) Service-Oriented Computing Fall

8 Restrictions Conceptually A unique feature of description logics Analogous to division in arithmetic: define classes in terms of a restriction that they satisfy with respect to a given property Anonymous: typically included in a class def to enable referring them Key primitives are somevaluesfrom a specified class allvaluesfrom a specified class hasvalue equal to a specified individual or data type mincardinality maxcardinality cardinality (when maxcardinality equals mincardinality) Munindar P. Singh (NCSU) Service-Oriented Computing Fall

9 Examples of Restrictions: 1 <owl : R e s t r i c t i o n > <owl : o n P r o p e r t y r d f : r e s o u r c e= #h a s F a t h e r /> <owl : m a x C a r d i n a l i t y r d f : d a t a t y p e= xsd : n o n N e g a t i v e I n t e g e r > 1 </owl : m a x C a r d i n a l i t y > </owl : R e s t r i c t i o n > <owl : R e s t r i c t i o n > <owl : o n P r o p e r t y r d f : r e s o u r c e = #bakes /> <owl : somevaluesfrom r d f : r e s o u r c e = #Bread /> </owl : R e s t r i c t i o n > Munindar P. Singh (NCSU) Service-Oriented Computing Fall

10 Examples of Restrictions: 2 The maker of a Wine must be a Winery The allvaluesfrom restriction is on the hasmaker property of this Wine class (Makers of other products such as cheese are not constrained by this local restriction) <owl : C l a s s r d f : ID= Wine > <r d f s : s u b C l a s s O f r d f : r e s o u r c e= &food ; P o t a b l e L i q u i d /> <r d f s : s u b C l a s s O f > <owl : R e s t r i c t i o n > <owl : o n P r o p e r t y r d f : r e s o u r c e= #hasmaker /> <owl : a l l V a l u e s F r o m r d f : r e s o u r c e= #Winery /> </owl : R e s t r i c t i o n > </ r d f s : s u b C l a s s O f >... </owl : C l a s s > Munindar P. Singh (NCSU) Service-Oriented Computing Fall

11 Axioms Conceptually Assertions that are given to be true Can be especially powerful in combination with other axioms, which may come from different documents Some primitives rdfs:subclassof owl:equivalentclass Munindar P. Singh (NCSU) Service-Oriented Computing Fall

12 Examples of Axioms <owl : A l l D i f f e r e n t > <! i n e s s e n c e, p a i r w i s e i n e q u a l i t i e s > <owl : d i s t i n c t M e m b e r s r d f : parsetype = C o l l e c t i o n > <ex : Country r d f : ID= I n d i a /> <ex : Country r d f : ID= R u s s i a /> <ex : Country r d f : ID= USA /> <owl : d i s t i n c t M e m b e r s/> </owl : A l l D i f f e r e n t > <ex : Country r d f : ID= I r a n /> <ex : Country r d f : ID= P e r s i a > <owl : s a m e I n d i v i d u a l A s r d f : r e s o u r c e = #I r a n /> </ex : Country> Munindar P. Singh (NCSU) Service-Oriented Computing Fall

13 Restrictions versus Axioms Axioms are global assertions that can be used as the basis for further inference Restrictions are constructors A restriction on hasfather of maxcardinality of 1 Does not mean all animals have zero or one fathers Means the class of animals who have zero or one fathers: this class may or may not have any instances Often, to achieve the desired effect, we would have to combine restrictions with axioms (such as based on equivalentclass), e.g., A restriction on hasfather of maxcardinality of 1 An axiom asserting this restriction is equivalent to Animal Munindar P. Singh (NCSU) Service-Oriented Computing Fall

14 Inference Like RDF, OWL is about meaning, not syntax Statements from different documents about the same URI are automatically conjoined OWL can be surprising to the uninitiated Integrity constraint: no one can have more than one mother Declare a fact: Mary is John s mother Declare a fact: Jane is John s mother What will you conclude? A traditional DBMS would declare an integrity violation An OWL reasoner would say Mary = Jane Munindar P. Singh (NCSU) Service-Oriented Computing Fall

15 Dialects Compared OWL DL Core dialect, includes DL primitives Not necessarily (but often) tractable OWL Lite Limits OWL DL constructs to ensure tractability No disjointwith, complementof, unionof, hasvalue Enumeration (oneof) intersectionof only for two or more class names or restrictions equivalentclass: class names to names or restrictions rdfs:subclassof: class names to names or restrictions allvaluesfrom and somevaluesfrom: to class names or datatype names rdf:type: to class names or restrictions rdf:domain: class names rdf:range: to class names or datatype names OWL Full Extremely general: allows all RDF syntax Potentially intractable (undecidable) Supports fancy expressiveness needs and introducing new concepts into the standard

16 Expressiveness Limitations: 1 OWL DL cannot express some simple requirements Non-tree models: because instance variables are implicit in OWL restrictions, OWL cannot express conditions that require that two variables be identified Think of siblings two people who have the same parents but in terms of classes Do the same thing with class definitions Munindar P. Singh (NCSU) Service-Oriented Computing Fall

17 Expressiveness Limitations: 2 Specialized properties Cannot state that the child of a mammal must be a mammal and so on, without Defining new child properties for each class Adding an axiom for each class stating that it is a subclassof the restriction of haschild to itself Analogous to the problem in a strongly typed object-oriented language without generics You have to typecast the contents of a hash table or linked list Munindar P. Singh (NCSU) Service-Oriented Computing Fall

18 Expressiveness Limitations: 3 Constraints among individuals and defeasibility Constraints among individuals Can define ETHusband: class of persons who have been married to Elizabeth Taylor Cannot define tall person: class of persons whose height is above a certain threshold Cannot capture defeasibility (also known as nonmonotonicity) Birds fly Penguins are birds Penguins don t fly Munindar P. Singh (NCSU) Service-Oriented Computing Fall

19 OWL Summary OWL builds on RDF to provide a rich vocabulary for capturing knowledge Synthesizes a lot of excellent work on discrete, taxonomic knowledge representation Fits well with describing information resources a basis for describing metadata vocabularies Critical for unambiguously describing services so they can be selected and suitably engaged Munindar P. Singh (NCSU) Service-Oriented Computing Fall

20 Modeling Exercise Student (S); faculty member (F); regular faculty member (R); department (D); thesis committee (T) An S belongs to exactly one D An R is an F An R advises zero or more Ss An F is affiliated with one or more Ds An S is advised by exactly one R An S is evaluated by exactly one T A T evaluates exactly one S A T has three or more Fs as its members Exactly one of the members of a T is its chair, who is an R Munindar P. Singh (NCSU) Service-Oriented Computing Fall