Re: [Proposal][SKOS-Core] skos:denotes

* Dave Reynolds <der@hplb.hpl.hp.com> [2004-09-29 16:56+0100]
> Dan Brickley wrote:
> 
> >SKOS is one layer of abstraction further removed from reality than 
> >basic RDF,
> 
> I guess I've seen it as another modeling style in parallel to 
> Class/Instance modeling rather than a layer more abstract.

In general, thesaurus-style and class/instance-style are at the same
level. But SKOS-in-RDF is an encoding of the former using the machinery of the
latter, which I guess is why it feels like a layering situation to me.

> >If several parties use SKOS and they have a concept in their
> >SKOS-expressed thesauri that stand for the person Alistair Miles, while
> >other parties simply write RDF statements about Alistair directly, we'd
> >imho benefit if we had some conventions for figuring out that they were 
> >talking about the same thing. But they're not "the same thing" in the
> >conventional RDF/OWL sense, since the class Person and the class
> >skos:Concept are presumably disjoint. 
> 
> Ah I hadn't realize they were expected to be disjoint. I nearly wrote 
> earlier that it you could simply have the bNode Al-as-foaf-Person also be 
> an instance of skos:Concept. Then it could, for example, be directly 
> attached to a thesaurus without this extra level of indirection and use 
> owl:sameAs to indicate these correspondences.

I think Alistair explored this at some point. Maybe it could be made to
work, though I fear things could get into a tangle, since the kinds of
metadata we like to have about people, and about thesaurus concepts, are
quite varied. Also, it would fold together concepts from different
SKOS thesauri that happened to denote/conceptualise/represent the same
things. So if we had workflow metadata like 'concept_creation_date', or
anything really, attached to indivdual concept nodes, we'd risk mangling
our data when things got merged.

In other words, we probably (imho) don't want to say that Alistair the
person is the selfsame thing as the representation of Alistair in the
SKOS from my weblog categories, and also as the representation of
Alistair in the SWED SKOS thesaurus (if he appeared there). It feels
more intuitive to model them as three distinct entities (which have 
different creators, for eg :), and take about
their inter-relationships. The SKOS mapping work allows us to do this
when talking about his appearance in several SKOS datasets; the new 
property would bridge this to the rest of the RDF world.

Am I making any more sense yet?

Dan

> 
> >Maybe skos:represents would work better as a name?
> 
> Yes, possibly.
> 
> Dave

Received on Wednesday, 29 September 2004 16:24:15 UTC