RE: [RIF] Extensible Design / Predicate Typing

> >>However, making the domain/range be a non-literal rdfs:Class 
> >>doesn't really impose any restriction.
> > 
> > It does, whenever most of your classes are disjoint, 
> > which is typically the case in any real world ontology. 
> 
> sorry for the later reaction on this thread, but doesn't this boil
> down to *querying* again?
> 
> i.e. in order to type-check an rdf typed variable, one needs 
> to *query*
> an OWL or RDFS ontology whether membership of the value bound to that 
> variable is entailed.
> 
> Thus, I do not exactly understand how:
> "3) Data literals, object names, function symbols
>      and predicate symbols may be typed. Using suitable
>      predicate/atom types, this allows to represent RDF
>      and OWL rules directly (and not only via a "query
>      interface")."
> 
> from an earlier mail shall be understood? ... Do you mean that this
> querying is implicit?

I was just pointing out that it is a different issue for a rule
language to be able to retrieve information from an RDF or OWL 
fact store by means of a "query interface" and to represent an
RDF or OWL rule for the purpose of interchange. The latter will
be required when we assume that there will be systems (such
as Jena or Pellet and Protege), which use an RDF or OWL/SWRL 
rule language, and RIF is supposed to support them. Since, for 
instance, SWRL has typed predicates/atoms (in fact, it has 7 
different types of predicates), but RuleML does not support 
this typing of predicates, RuleML cannot be used for interchanging 
SWRL rules, say, between a Pellet and a Protege system (and the 
same applies to Jena's RDF rules whenever property predicates 
are typed by means of rdfs:domain and rdfs:range).

-Gerd

Received on Monday, 22 May 2006 10:51:52 UTC