Re: SemWeb terminology page

On Fri, Dec 03, 2010 at 03:58:56PM -0500, Jeff Young wrote:
> Yes. (I'm happy to sidestep issues of "value" vs. "label" by assuming
> "lexical entity" is the common denominator.) SKOSXL allows these to be
> treated as identifiable resources. Also note the openness of the domain
> on skos:inScheme:
> 
> http://www.w3.org/TR/skos-reference/#L2805
> 
> This means that skosxl:Labels can be "controlled" using skos:inScheme
> without requiring a skos:Concept intermediary. (I admit this may be
> controversial, but the basic justification can be found here:)

I have long recognized this theoretical possibility, but that
I am aware, you are the first person to actually point out
that one _could_ define not a concept scheme, but in effect a
"label scheme" :-)

> It would be odd to dismiss SKOS because we determined it was designed to
> manage "concepts" rather than "controlled vocabularies".

I certainly wouldn't want to dismiss SKOS!  The point is that
SKOS organizes sets of lexical strings via underlying concepts.

> > In metadata, a concept in the LCSH concept
> > scheme would typically be cited in metadata using the URI of
> > the concept.
> 
> Agreed. And that concept has a skos:prefLabel that relates to the notion
> of a "value". In principle, LCSH could identify skosxl:Labels instead of
> (or in addition to) skos:Concepts. OTOH, I think we can prove that the
> concepts associated with LCCNs are much more stable than the
> labels/values. SKOS Concepts are also easier to relate across controlled
> vocabularies than SKOSXL Labels are.
> 
> > This is also in line what an earlier posting proposed as
> > a definition for Group 1: "1) LCSH, AAT, WordNet and the
> > like. These describe concepts that are used in actual medata."
> 
> How about some fuzzy wording like this:
> 
> These provide consistent names/labels for conceptualized things that can
> be used in actual metadata.

Hmm, I'd rather not go there...  I think it complexifies the
definition unnecessarily to focus on "consistent names" (though
I see this would apply for certain types of authority control).
Actually, I am not aware of anything in the SKOS data model
that _requires_ a SKOS concept to have a label at all -- and
I don't think we would want to exclude label-less concepts
from this category. In a way, names _are_ conceptual things,
in the sense that words are conceptual things.  As for the
definitions, I think shorter is better; if anything, a few
more illustrative examples might help.

Tom

-- 
Thomas Baker <tbaker@tbaker.de>

Received on Friday, 3 December 2010 22:00:10 UTC