FullTopicsRDFSemantics

OWL-1.1-Full Topic: RDF(S) Semantics

Issues in RDF(S)

Possible types of issues:

• BUG: An error
• FEATURE: A request for a new feature
• CHANGE: A request for changing a current feature

FEATURE: bNodes should be allowed in predicate position

ter Horst shows in the pD* paper that not having bNode predicates is a problem for the completeness proof of the entailment lemma (sec. 1.5), and that this problem can be removed by allowing such bNodes ("extended RDF graphs").

CHANGE: Annotation properties should have range rdfs:Resource

Currently, the following axiomatic triples exist in the RDFS spec:

• rdfs:comment rdfs:range rdfs:Literal
• rdfs:label rdfs:range rdfs:Literal

This is a problem, since people like for instance to have graphics as labels.

CHANGE: domain and range of all properties should be rdfs:Resource

Currently, AFAICS, RDFS doesn't provide the "obvious" fact that subject and object of a triple are resources. There are only several axiomatic triples, which specify the domain and range of the properties from the RDFS vocabulary specifically.

I suggest to

• remove all the axiomatic of the form p rdf:domain/range C,
• add axiomatic triples of the form "rdfs:comment rdf:type rdf:Property
• add the following semantic condition to RDFS:

 IF
     p in IP
 THEN
     (p, I(rdfs:Resource)) in IEXT(I(rdfs:domain))
   AND
     (p, I(rdfs:Resource)) in IEXT(I(rdfs:range))

This would automatically include all removed axiomatic triples.

Because of the "if" semantics of RDFS for rdfs:domain and rdfs:range, this is a stronger condition than

 IF
     p in IP
   AND
     (x,y) in IEXT(p)
 THEN
     x,y in ICEXT(rdfs:Resource)