70 likes | 203 Views
A Proposal for Addressing Issues Related to RDF Mapping. Outline. Disallow punning between object and data properties Require that each URI used as a property must be explicitly typed as either an object or a data property Modify all the related deliverables RDF mapping
E N D
Outline • Disallow punning between object and data properties • Require that each URI used as a property must be explicitly typed • as either an object or a data property • Modify all the related deliverables • RDF mapping • map declarations/typing axioms into rdf:type • get rid of duplicate vocabulary • XML syntax • functional-style syntax
Changes to Structural Specification • Move the discussion about declarations/typing axioms into the core part of the spec • Allow all entities to be declared • solves the “I want to put a class into an ontology” problem • Typing axioms are mandatory if there is ambiguity • object vs. data property • class vs. datatype • If a URI u is used as an object property in an ontology O then the import closure of O… • …must contain a declaration saying that u is an object property • …must not contain a declaration saying that u is a data or an annotation property • The situation is analogous for data and annotation properties
Changes to Functional-Style Syntax • Current: • objectPropertyURI := URI • objectPropertyExpression := objectPropertyURI | … • objectSomeValuesFrom := 'ObjectSomeValuesFrom' '(‘ objectPropertyExpression description ')‘ • Proposed: • objectPropertyURI := URI[OP] • objectPropertyExpression := objectPropertyURI | … • objectSomeValuesFrom := 'SomeValuesFrom' '(' objectPropertyExpression description ')‘ • Recommend: put all property typing statements at the top of the file Means that the import closure of the ontology declares URI as an object property
Changes to RDF Mapping • Map declarations into rdf:type • e.g., a declaration that u is an object property is mapped into < u, rdf:type, owl:ObjectProperty > • Get rid of duplicate vocabulary • solves the problem of nonmonotonocity of the mapping • State that, in the reverse transformation, only explicit rdf:type triples count as declarations • the triples implied by the RDFS semantics can be used for repair, but, strictly speaking, do not constitute declarations • Recommend: put all property typing statements at the top of the file
RDF Mapping – Example • OWL 1.1 FS ontology: • Declare(OWLClass(A)) • Declare(OWLClass(B)) • SubClassOf( A B ) • The RDF mapping will be: • < A, rdf:type, owl:Class > • < B, rdf:type, owl:Class > • < A, rdfs:subClassOf, B > • Currently, you also got the following triples, but this will go away • < A, owl:declaredAs, owl:Class > • < B, owl:declaredAs, owl:Class >
Changes to XML Syntax • Similar to changes in the functional-style syntax • drop ObjectSomeValuesFrom in favour of SomeValuesFrom