On 2007 Oct 8, at 18:09, Bernard Vatant wrote:
> So something like the following would be perfectly consistent with
> both OWL and SKOS semantics
>
> iau:stars rdf:type skos:Concept
> cds:stars rdf:type skos:Concept
> iau:stars owl:sameAs cds:stars
>
> Of course, you might want things more subtle than that, because for
> some reason you want to declare a mapping without merging the two
> concepts. In that case you can use SKOS mapping properties, but
> note that this draft has no formal status whatsoever so far on W3C
> track. See http://www.w3.org/2004/02/skos/mapping/spec/
In the demo I gave at the ADASS VOEvent BoF, I illustrated how you could get this same effect using rdfs:subClassOf:
<ivo://VOcabulary/AOIM#startypevariablenova>
a skos:Concept;
skos:broader <ivo://VOcabulary/AOIM#star>;
rdfs:subClassOf <http://www.ivoa.net/ns/ucd#star.binary.CV.nova> .
<ivo://ivoa/VOcabulary/AAkeys#starsnovaecataclysmicvariables>
a skos:Concept;
skos:broader <ivo://ivoa/VOcabulary/AAkeys#stars>;
rdfs:subClassOf <http://www.ivoa.net/ns/ucd#star.binary.CV.nova>.
<http://www.ivoa.net/ns/ucd#star.binary.CV.nova>
rdfs:subClassOf <ivo://VOcabulary/AOIM#startypevariablenova>;
rdfs:subClassOf <ivo://ivoa/VOcabulary/
AAkeys#starsnovaecataclysmicvariables> .
(the pairs of subClassOf relations mean you can go both ways). That meant that the SPARQL query
select ?r ?b
where {
AOIM:startypevariablenova rdfs:subClassOf ?r.
?r skos:broader ?b.
}
was able to start from the AOIM vocabulary and find a broader term in the A&A vocabulary, using only standard tools.
That has the effect of owl:sameAs[1], but using only an RDFS reasoner.
Rick on 2007 October 9:
> It's the "sameAs" functionality which we need to be able to
> translate between vocabularies.
We can support a mixture of such direct pairwise mappings, or else go via an intermediate vocabulary such as UCDs
> This is all very interesting (seriously!), but it should be clear
> that there's no way we're going to be able to produce a version of
> the IAU/IVOA where the ontological hints have been clean up /
> checked. We should consider the present content to be informative
> and condusive to testing ontological analyses with thesauri. The
> goal might be to go back and re-check the BT's, NT's, and RT's and
> maybe even add more info in the next version (or an auxilliary
> version).
Once it's got into a triple store, it doesn't matter where RDF has come from. Thus it's perfectly practical to have a SKOSified vocabulary as one deliverable, and a set of mappings and additional annotations in a separate auxiliary deliverable.
All the best,
Norman
[1] perhaps modulo some class-as-instance subtlety
-- ------------------------------------------------------------ Norman Gray : http://nxg.me.uk eurovotech.org : University of Leicester, UKReceived on 2007-10-10Z12:05:16