On 23/2/09 20:04, Danny Ayers wrote:
2009/2/23 Matthias Samwald<[email protected]>:
In
contrast, I would still prefer an established, generic property to something
that needs to be invented anew. rdfs:seeAlso is established (which also
means that people are more likely to re-use it then something you create in
your own little namespace), and it is also understood by some existing
applications. If all of us are starting to create our own "is quite similar
to" predicates, we are probably worse off than just using rdfs:seeAlso. The
semantics of these "is quite similar to" relations would probably be so
fuzzy that they would hardly be any more precise and useful than
rdfs:seeAlso, anyways.
There does seem to be a considerably more precise correspondence
between Cyc concepts and WordNet terms than rdfs:seeAlso describes.
But taking your point of what existing applications understand,
perhaps it might make sense to use rdfs:seeAlso in parallel with a
custom term?
That was to be my original suggestion, but I trimmed the sentence about
rdfs:seeAlso as I was too lazy to defend the suggestion.
There is certainly much richer information here than is captured by
rdfs:seeAlso. Furthermore, it is a very useful kind of information: that
which associates the core machinery of RDFS/OWL (classes) with the
natural language vocabulary used by mere mortal humans.
I'm not sure that the wordnet: Synset class is the final say in
linguistics-related RDF vocab, but some bridging property from classes
to this makes a lot of sense. In a SKOS concept I have been revisiting
the idea of an "it" property that relates a SKOS concept to "the thing
itself". If the Synset were modelled as a SKOS concept, then the class
here is "the thing itself". See
http://www.flickr.com/photos/danbri/3282565132/ for the "it" business. I
bounced this idea off Alistair Miles (putting "it" into FOAF) and he
thought it fine, since the SKOS group didn't handle this within SKOS.
I'm interested if this might be another use case for such a property...
Dan
