On Sep 17, 2009, at 3:54 PM, Sean Bechhofer wrote:
>
> Absolutely. Working with the underlying RDF graph rather than its
> concrete serialization is (IMO) the right thing to do, and SPARQL,
> rather than XSLT will give you that.
XSLT is a hammer from an alternate universe where the fundamental
particle of matter is the thumb :-P
In most cases, there is a Class whose instances correspond to the
Things which would be mentioned in a Doculope* labeled with a Concept;
however the Concepts and the Classes usually want to be different
entities, and the mapping seem to be easier to describe in SWRL-DL
rather than OWL. This is a situation where distinguishing BTG and BTP
separately helps a lot. BTI can be handled trivially using roles, as
long as the end point of the chain is Object, not Data.
I've been trying this for KOSS (using Attempto Controlled English) to
try and get more precise semantics for precombined and subdivided LCSH
headings. Things seem to make more sense if you treat the inferred A-
box as a T-box for an actual ontology. This makes a lot of sense if
you treat the domain of interpretation of SKOS as the Doculopes, not
the underlying Things.
Simon // Back in England for the high holidays. Isn't it still BBQ
summer?
* Doculopes are documents a la Buckland.