- From: Hugh Glaser <hg@ecs.soton.ac.uk>
- Date: Wed, 15 Oct 2008 10:57:32 +0100
- To: Bernhard Schandl <bernhard.schandl@univie.ac.at>, Knud Hinnerk Möller <knud.moeller@deri.org>
- CC: Thomas Loertsch <loertsch.thomas@guj.de>, Semantic Web <semantic-web@w3.org>
[Hit the wrong key, please ignore previous email] Not addressing performance, but more the advice question, and linkage and re-use: Yes, rdfs:subClassOf and rdfs:subPropertyOf seem more sensible than owl:equivalentClass for many things, and certainly for rdfs:label sorts of things. If you can use owl:equivalentClass, why did you invent a new concept and not use rdfs:label in the first place? We find ourselves doing things like rdfs:subPropertyOf rdfs:label for many things when we use stuff from outside, so that our human interface tools can work. It would be a great help if providers could do that for us, and after all they are the people who know best what is sensible. So Thomas says he would like to see guidance such as: "If you want to express that something has a name/title/label, use rdfs:label." Such advice might be enhanced by: If you have something that can play that role, then please establish appropriate linkage to rdfs:label in the following way. This seems to be very much the Linking Data world, but at the ontology level, which is where some focus is now being turned, with the new diagram appearing on http://esw.w3.org/topic/SweoIG/TaskForces/CommunityProjects/LinkingOpenData. Giving some more guidance on best practice, rather than allowing complete organic growth is often good. Best Hugh On 15/10/2008 07:07, "Bernhard Schandl" <bernhard.schandl@univie.ac.at> wrote: > > > Hi Knud, > >>> That way a query for everything with a 'title' could automagically >>> sparql >>> the clearinghouse for every owl:equivalentClass of 'title' and add >>> those to >>> the original query. That's a scalable solution and it doesn't need >>> much >>> (actually none if I'm not mistaken) inferencing power either. >> >> That's an interesting idea. It would be interesting to find how this >> compares performance-wise to "ordinary" inferencing. > > We did this regularly in several applications, however we were not > using owl:equivalentClass but rdfs:subClassOf and rdfs:subPropertyOf; > and we applied this not only to labels but also to class instances: > > SELECT ?r ?l > WHERE > { > ?r rdf:type ?t . > ?t rdfs:subClassOf my:Something . > ?r ?p ?l . > ?p rdfs:subPropertyOf rdfs:label . > } > > This of course requires the full closure of subClassOf and > subPropertyOf relationships to be computed (but this can usually be > done once and then is persisted). We found this performs quite well in > comparison to fully fledged inferencing; however it has the drawback > that reasoning is encoded into applications, which one might find not > a nice solution. > > Best regards, > Bernhard > > >
Received on Wednesday, 15 October 2008 09:59:11 UTC