- From: Pierre-Antoine Champin <pierre-antoine.champin@liris.cnrs.fr>
- Date: Thu, 26 Aug 2010 10:31:28 +0200
- To: "Evain, Jean-Pierre" <evain@ebu.ch>
- CC: Tobias Bürger <tobias.buerger@salzburgresearch.at>, "public-media-annotation@w3.org" <public-media-annotation@w3.org>
On 26/08/2010 07:01, Evain, Jean-Pierre wrote: > 1/ Actually, I had another look at the RDF specification for the use > of UnionOf. I don't think it applies here. which one? the 1999 specification was bugged in that respect http://www.w3.org/2000/03/rdf-tracking/#rdfs-domain-and-range > Let' take another example... > > We have the property 'location' (where a resource can be accessed > from, typically a URI). A locator can be assocaited to a media > resource, a fragment or a named fragment. the important word here is "or" ! If you state the following triples: ma:locator rdfs:domain ma:MediaResource . (1) ma:locator rdfs:domain ma:Fragment . (2) then each of those triples can be considered independantly of the other. Now consider :res1 ma:locator :loc1 . (3) >From (1) and (3) I can infer that :res1 is a ma:MediaResource. >From (1) and (2) I can infer that :rest1 is also a ma:Fragment. In other words, you have stated with (1-2) that a locator can be associated to (something that is) a media resource *and* a fragment. The only way to express the "or" in your sentence above is to use a owl:unionOf . > Therefore it make sense to > say that the property location is multirange. It will be used to > generate valid clearly separated triples. I can't see the use of > UnionOf here. Or? I agree that aggregating all this in a unionOf construct is a bit awkward, but this is required: separate triples have a *conjunctive* semantics. pa
Received on Thursday, 26 August 2010 08:32:05 UTC