Search Mailing List Archives
[protege-owl] Tutorial for using SPARQL with OWL.
holger at knublauch.com
Wed Sep 6 11:15:40 PDT 2006
> Gerhard's question does not mention asserted vs. inferred.
> I took it as a question about retrieving the "indirect" instances of a
> class, i.e. the instances of some subclass of a class, no matter if the
> subclass is asserted or infered.
But indirect instances have an inferred rdf:type triple. If you have
reasoning turned on on a Jena graph and then ask for the types of a
resource, then it will return all types (up to rdfs:Resource if you
want) as extra, inferred triples.
> Again, I don't think that this is possible because sparql is not aware
> of the rdfs semantics (hence sparql is not aware of owl's semantics).
> sparql doesn't understand the subclass assertions (which are processed
> as any assertion) and it doesn't understand that rdfs:subClassOf is
> transitive, which is also a problem in the case proposed by Gerhard.
But these inferences can be handled further down on the Graph level. If
you have a Graph implementation that does the inferencing on the fly
then you can execute SPARQL on top of it. You can get complete OWL DL
semantics by taking a Jena Pellet Graph as the query base.
> Other rdf query languages are aware of the rdfs semantics (e.g. SeRQL).
> Therefore, the fact that Gerhard's wish is impossible is not a
> limitation of Protege but only of sparql. Holger's remark is
> perfectly correct, but gives the impression that the problem is Protege
> whereas it is sparql (I think).
Nope, it's a Protege problem only.
More information about the protege-owl