On Jul 29, 2008, at 3:42 PM, Bernard Vatant wrote:
[snip]
Recommending systems built on the semantics of OWL, which are quite
hard to set up and define properly, to ignore parts of this very
semantics sounds to me as a strange recommendation coming from the
very W3C top :-)
But : as said before, the real issue is that you ask both
vocabulary publishers and implementers to hack the standards
semantics, default any proper way in those standards to deal with
names.
What I suggest here (with cc to SKOS forum) is to define somewhere
a standard generic label datatype property, which could be easily
refined in specific types of labels while keeping in OWL-DL; making
everybody happy. The most natural place to do that seems to be SKOS
namespace.
Here is the proposal
skos:label a owl:DatatypeProperty
skos:prefLabel rdfs:subPropertyOf skos:label
skos:altLabel rdfs:subPropertyOf skos:label
...
pub:name rdfs:subPropertyOf skos:label
[snip]
I think considering a nice set of label properties is a good idea.
I'd like it if we considered various mechanisms for describing them
as well as opposed to the subproperty trick. (Some variant of which,
of course, can be made to work in OWL esp. in OWL 2. I'm happy to
push for extra things in OWL 2 to accommodate this better.)
(For example, one thing that seems like it would be really useful is
to designate fallbacks. alt:label seems like one, but wouldn't it be
better to be able to say something like: display the label, if
present, then the altLabel, if these are missing then use the string
after the hash of the uri or the content of the property "name".)
(Given the commonality of certain uri naming conventions, we might go
even further. E.g., define certain parsing for hasFoo.)
Cheers,
Bijan.
