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

Can open - worms everywhere (Chandler Byng which is a bit sad)!

As good scholastics, when in dispute, appeal to ancient authority.
E.g. Doerr+Fundulaki:
http://www.ics.forth.gr/isl/publications/paperlink/A_proposal_on_extended_in
terthesaurus_links_semantics.ps.gz

says:

"...considering a concept as a reference notion by which some people 
agree to classify a set of real world objects having the same properties,
but without necessarily defining these properties. ... In this context we 
perceive a concept as a surrogate for the objects correctly classified 
under the intended aspects of the thesaurus maintainers."

Later in the same paper, they say that under this interpretation:

"if Descriptor_A has a broader meaning than Descriptor_B then the instance
set is of the latter is a subset of that of the former."

As I interpret this,  this gives the best semantics I know of for 
Thesaurus concepts - a concept's denotation is "the set of resources 
which are classified under that concept" (my words).  This is quite 
different from the intended use in Ontologies, where instances are 
"supposed" to stand for "real-world" things (with suitable philosophical 
quotation marks).

Concepts are thus in a sense not intepretable (they don't have properties
in the ontology sense).  Note the change of levels; Al's skos:denotes 
is intended to often relate an instance of a class of concepts to an 
RDFS/OWL class.

Personally, I think that the term Concept in thesaurus is misleading;
a better word would be Topic or Subject (we already have dc:subject 
proposed for use in skos).  However, Concept is deeply ingrained
in the thesaurus culture, and Topic (Topic Maps) and Subject (Natural
Language,
RDF) have other implications, so we should stick with Concept.

So I think that the word "denotes" to connect a thesaurus-theoretic and
a ontology-theoretic point of view is dangerous - as in logic
and mathematics this usually signifies semantics - we are not (should
not) be saying that the class provides a semantics for the concept.

An alternative term then?  skos:classifierFor is perhaps most precise,
though a bit awkward.  But I prefer it to the alternatives we have
seen so far which imply some kind of semantic relationship.

I do agree with Nikki that I want to see the convincing use case

Nikki says:
> Or is this debate really about the fact that we want to stick some 
> machinery capable of SKOS<->OWL stuff in SKOS-Core right now, so that 
> SKOS-Core stands in its own right allowing us to tackle SKOS-Mapping 
> separately?

As far as I see - yes!  The use of this property is to do the modelling
task of relating SKOS and OWL - and SKOS-mapping should be separate.

To turn to the specifics of the proposal (to start a new controversy),
does it make sense (particularly when taking a point of view that
we are providing a classifier for an RDF resource) for this property
to be a functional property?  This would mean if two resources have the
same classifier, they can be identified.  This may not make sense.

Brian

> -----Original Message-----
> From: public-esw-thes-request@w3.org
> [mailto:public-esw-thes-request@w3.org]On Behalf Of Dave Reynolds
> Sent: 29 September 2004 17:26
> To: Miles, AJ (Alistair) 
> Cc: Dan Brickley; 'public-esw-thes@w3.org'
> Subject: Re: [Proposal][SKOS-Core] skos:denotes
> 
> 
> 
> Miles, AJ (Alistair) wrote:
> 
> >>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.
> > 
> > 
> > The big problem with doing this is that multiple concepts 
> from different
> > sources could end up being merged as the same node in a 
> graph ... which
> > could leave you with a concept node with three different 
> definitions and any
> > number of labels, and no knowledge (without provenance) of 
> which source the
> > labels/definitions came from.
> 
> Well you often need provenance to work with merged data 
> anyway but I agree 
> there are advantages to keeping the chunks separate rather 
> than relying on 
> provenance machinery.
> 
> However, it seems like the core argument behind the approach 
> is not the 
> mechanics of merging but the nature of modeling.
> 
> All of these representation systems are approximations. A 
> skos Concept and 
> an rdfs Class may both capture different nuances about the 
> same thing [*]. 
> You want some way of saying that. It can't carry any 
> semantics but it is 
> useful. However, in that case it seems to me the relationship is 
> "denotesTheSameAs" which is why your phrasing of 
> "correspondence" works 
> better for me than denotes. In particular, this seems to me 
> to be quite 
> symmetric. You might want to say that two different skos 
> Concepts from 
> different thesauri also denote the same thing. It need not be 
> a specific 
> SKOS -> RDFS/OWL mapping. I think it was the asymmetry that I 
> didn't like.
> 
> Dave
> 
> [*] I would use "real world thing" here but that would just 
> get it confused 
> with that subclass of things that hurt your toe when you kick them ...
> 
>

Received on Thursday, 30 September 2004 11:18:41 UTC