- From: Bernard Vatant <bernard.vatant@mondeca.com>
- Date: Tue, 07 Nov 2006 17:50:20 +0100
- To: Mikael Nilsson <mini@nada.kth.se>
- Cc: "Williams, Stuart (HP Labs, Bristol)" <skw@hp.com>, public-esw-thes@w3.org
Mikael > Well, apart from the fact that I agree with most things you've said Which is good news, thanks. > I DON'T think this way of describing things is useful for all kinds of > RDF. I only mean it to apply to cases where you do explicit > conceptualizations, such as in concept schemes. > Including ontologies, I guess. I suppose in other cases you make implicit conceptualizations. :-) I don't think you can use RDF or any description language without sort sort of conceptualization. You don't handle things in languages, only their description (that sounds terribly obvious, sorry.) > For example, if I have the two nodes _:me and _:mycar in a simple > Person/Car ontology, I see no need to complicate things - _:me can well > be owl:sameAs any other node that represents me. > Yes. Don't take me wrong : I don't propose to discard owl:sameAs altogether. Those are cases where you want to indeed merge information, without the subtleties of viewpoints, history, and and other flavors of description. But beware of "any other node that represents you". Even if you are not explicitly conceptualized, so to speak, this will bring inconsistency very soon, if you put properties such as age, job, number of children, whatever is some element of description of you, but is bound to change (that is, almost anything). > But in cases where *the conceptualization itself* is described and has a > history, we need the distinction. > This is indeed a clear use case to begin with, versioning and the like. But seems to me the scope of use is wider than that. > or so I think :-) > And so do I ;-) Bernard > /Mikael > > >>> It seems to me that skos:Concept is an *explicit* conceptualization of >>> some thing. With a history, purpose, etc in itself, separately from the >>> thing. >>> >>> >> Agreed. So you need this level of indirection between the thing and the >> resource which describes it. >> >>> so by using owl:sameAs, we're saying that not only are the described >>> "things" the same, but we're also using the *same* conceptualization, >>> with the same history etc. >>> >>> >> Absolutely. So the two descriptions/resources are merged, and that's not >> what we want. >> >>> OTOH, it's certainly an interesting statement to say, well, we're >>> talking about the same thing (subjectIndicator) but we're using >>> different conceptualizations, etc. >>> >>> >> Certainly. But the point of Stuart is that if subjectIndicator is an >> IFP, the logical result is the same (sorry) as if you use directly >> owl:sameAs >> >>> So, in short, couldn't the answer be that we are really talking about >>> two resources - the thing and the conceptualization? >>> >>> >> ABSOLUTELY! I'm happy you come to this conclusion at the end of this >> kind of Socratic dialogue. Now all my point with "hubjects" or "blank >> subjects" is that the resource which is the thing is beyond any >> description - otherwise you get into a recursivity loop. And subject >> indicator is not a killer solution to that, as I discovered after >> passing years munching this notion in OASIS Technical Committtee. The >> subject indicator is yet another conceptualization for the thing (less >> formal, more for humans, but the issue remains the same). >> >> That's why I suggest this thing beyond the >> conceptualizations/descriptions to be pointed as a really blank node, >> using whatever relevant pointer. I have discussed this with Tom Baker >> who did not see any formal opposition to use dc:subject here, but it's a >> quite weird use of it. I think we really need a specific property, and >> of course NOT an IFP. >> And I would be happy to have it in skos namespace, something like >> skos:isConceptFor, so we would have >> >> a:thisConcept skos:isConceptFor _:thingFoo >> b:thatConcept skos:isConceptFor _:thingFoo >> >> The _:thingFoo balnk node having no other purpose that linking the two >> concepts without merging them, and of course no rdf:Description whatsoever. >> >> Well, I think I've hit that nail more than enough for now. >> >> -- *Bernard Vatant *Knowledge Engineering ---------------------------------------------------- *Mondeca** *3, cité Nollez 75018 Paris France Web: www.mondeca.com <http://www.mondeca.com> ---------------------------------------------------- Tel: +33 (0) 871 488 459 Mail: bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com> Blog: Leçons de Choses <http://mondeca.wordpress.com/>
Received on Tuesday, 7 November 2006 16:50:33 UTC