RE: [Proposal][SKOS-Core] Moving semantic relation property exten sions to an 'extensions' vocab.

I thought it might be worth saying a word or two on this ...

Many moons ago (when I was barely a babe in arms) I imagined that there
could be a seamless continuum between thesauri and ontologies, mediated via
RDF, and exploiting the semantics of existing RDF RDFS and OWL terms.  

I.e. you could start off with a 'thesaurus-style' description of a set of
resources, and then by simpling adding statements, end up with an ontology.


This idea was behind the description of e.g. 'skos:broaderGeneric' (which
has remained unchanged since first introduced): you can see that this term
is a sub-property of both 'skos:broader' and 'rdfs:subClassOf'.  This means
that, at the moment, if somebody states (X skos:broaderGeneric Y), that
implies both the fuzzy thesaurus style (X skos:broader Y) and the ontology
style statement (X rdfs:subClassOf Y).

Likewise, 'skos:broaderInstantive' is a subprop of both 'skos:broader' and
'rdf:type'.

Then we had a bit of a discussion over a SWAD-E meeting one afternoon, and
it was suggested that actually it is better to keep the two worlds apart -
i.e. you describe a set of *thesaurus* resources using *thesaurus*
predicates, and you describe a set of *ontology* resources using a set of
*ontology* predicates.  The two sets of resources could then be linked via a
special predicate, which we couldn't decide what to call, but is the origin
of danbri's reference to a 'denotes' predicate I think (danbri am I right
there?)  Another suggested name was 'conceptualises' ... i.e. 'thesaurus
resource A *conceptualises* ontology resource B'. 

Anyway, the point I want to make is: this whole issue looks like it needs a
lot of discussion.  And the original reasoning behind the 'extension'
properties (i.e. skos:broaderGeneric etc.) may need to be completely
reconsidered.  This is the main reason I proposed to remove them from the
SKOS Core vocab and into a separate 'extensions' vocab for now - so we can
publish an improved SKOS Core guide and spec in the short term, without
getting into a possibly sticky debate about thesaurus -> ontology that may
not get resolved for a while.

Al.



---
Alistair Miles
Research Associate
CCLRC - Rutherford Appleton Laboratory
Building R1 Room 1.60
Fermi Avenue
Chilton
Didcot
Oxfordshire OX11 0QX
United Kingdom
Email:        a.j.miles@rl.ac.uk
Tel: +44 (0)1235 445440



> -----Original Message-----
> From: Dan Brickley [mailto:danbri@w3.org]
> Sent: 22 September 2004 10:00
> To: Miles, AJ (Alistair) 
> Cc: 'public-esw-thes@w3.org'
> Subject: Re: [Proposal][SKOS-Core] Moving semantic relation property
> exten sions to an 'extensions' vocab.
> 
> 
> * Miles, AJ (Alistair)  <A.J.Miles@rl.ac.uk> [2004-09-21 18:38+0100]
> > 
> > Any further comments on this proposal?
> > 
> > 
> <http://lists.w3.org/Archives/Public/public-esw-thes/2004Aug/0
> 081.html>
> > 
> > ... 
> > 
> > I would like to propose that the following properties be 
> removed from 
> 
> seconded!
> 
> > SKOS Core, and be moved to an 'extensions' vocabulary 
> (perhaps under the 
> > namespace <http://www.w3.org/2004/02/skos/extensions#>?) ...
> > 
> > skos:broaderGeneric
> > skos:broaderInstantive
> > skos:broaderPartitive
> > skos:narrowerGeneric
> > skos:narrowerInstantive
> > skos:narrowerPartitive
> > skos:relatedHasPart
> > skos:relatedIsPartOf
> > 
> > My reasons are:
> > 
> > (1) These are the least stable parts of SKOS Core, and I 
> don't expect 
> > them to stabilise in the short term (i.e. months).  So 
> they're getting 
> > in the way of publishing short term.
> > (2) They impinge on the whole 'thesauri -> ontologies' 
> question, which 
> > again I don't think we're going to have an answer for in 
> the next couple 
> > of months.
> > (3) They clutter up SKOS Core, and distract from its 
> fundamental features.
> 
> Yup. They're interesting, but awkward. These properties are 
> in the area 
> in which RDF's (well, OWL's) facilities overlap with the more
> sophisticated things that people have tried doing with 
> thesauri. Facets etc.
> 
> So there's a risk of reinventing some of OWL, but at one layer of
> abstraction's remove, and without the classes-and-instances conceptual
> model that allows OWL to be mathematically precise about its 
> notion of 
> hierarchies of restriction.
> 
> I certainly think it's worth exploring, perhaps in the context of
> guidelines for migrating thesauri into fully-fledged RDF/OWL vocabs.,
> but I definitely support moving this out of the core.
> 
> (that'd make room in the core for a 'denotes' relation, hmm, where did
> that discussion get up to...?)
> 
> cheers,
> 
> Dan
>

Received on Thursday, 23 September 2004 16:36:47 UTC