Re: [VM,ALL] Revised scope statement

On Thu, Jun 17, 2004 at 11:30:02AM +0200, Bernard Vatant wrote:
> > To my way of thinking, the Vocabulary Management TF would,
> > in contrast, focus on the identification of terms (and of
> > versions of terms and sets of terms) and on policies and
> > practices related to the identification of terms.
> 
> The real issue here is to know if it makes sense to identify terms (or anything else)
> independently of any application context. For example, in a Thesaurus, the application
> context of a term (i.e. its contextual definition) is expressed by its BT, NT, RT, UF,
> USE, Scope Note ... If you strip a term off all this contextual information, what's left?
> a name? a URI? A bare identifier without any identification context is as useful as a
> credit card number outside any banking system.

What you refer to as "contextual information" -- BT,
NT, etc... -- is what I would call "semantic context".
The thesaurus maintainers are using these relators to assert a
semantic context for their own term.  

To my way of thinking, this semantic context is different from
the identification context per se.  If a URI identifying a
term is backed up by policies and explanations formulated by
the coiners and maintainers of that URI, then to me that URI
is not simply a "bare identifier".  Rather, the maintainers
are be establishing for that URI an "identification context".

Though the richer semantic context asserted for a term
by its maintainers (i.e., the sum of its relators) must
be understood in order to use the terms as intended, the
"bare identifier" does not itself reflect or carry this
semantic context.  Once a URI is used to identify something,
it can in principle be used in SW assertions.  In principle,
"anyone" could assert a different or even a semantically
perverse context for someone else's thesaurus term, but that
does not mean one cannot or should not make such assertions.

> IOW, relationships between identification and contextual definition are tricky to
> entangle, and setting generic term identification valid for *any* context seems very
> difficult (read : barely possible).

In my understanding, the whole idea of Semantic Web is
based on the notion of citing and reusing and recombining
and referencing and repurposing and merging data from a
diversity of sources, and it is well recognized that doing
so can in principle violate the semantic intent of the
source data.  To question whether it ever makes sense to use
terms independently of their application context would seem
to question the very idea of Semantic Web.  I don't think
that is what you are saying, but I do think one can and must
distinguish between how maintainers "make sense" for a term
(which I see as largely out of scope for the VM note) and
what simply establishes identity and makes terms referencable
by others, "sensibly" or not.

> > But that's okay -- at that point, the VM TF could simply point
> > off to other documents and practices such as the the THES/PORT
> > TF note and the OASIS Published Subjects work you cite below.
> 
> If that means : There is a generic question of term identification, generic principles
> that can be set in the SW context (see below), but specific ways to apply those principles
> always depend on context (e.g. Thesaurus, Ontologies, Topic Maps, Taxonomies ...) then I
> agree.

Yes.

> BTW such an approach could help to get out of the endless debate on URI meaning, by
> stressing the (IMO obvious) fact that whatever an identifier identifies necessarily always
> assumes an application context, and that the Web (semantic or otherwise) can barely be
> considered a univocal application context ...

Yes!  I should have read your whole message before starting my reply.

> > > 2. I share the concern expressed by Alan about "terminological" vs
> > "conceptual" approaches
> > > of Vocabulary, and the need for clarification about it in the SW community.
> > SKOS input is
> > > certainly to be brought to the table, as well as current debates about use of
> > dc:subject
> > > in various places.
> >
> > My instinct would be to cite such debates where appropriate
> > but to put alot of these issues out of scope for the VM TF
> > note itself and focus on lower-hanging fruit.  For example,
> > can we agree that terms should be both identified with URIs
> > and labelled with human language?
> 
> Hopefully this is a reasonable consensus basis.

Excellent.

> > I read it as
> > saying, in essence: "Subject headings intended for use with
> > Topic Maps should be identified with URIs, labelled with human
> > language, accompanied with a statement of intended use, and
> > described with metadata."
> > If this paraphrase does justice
> > to the recommendation, then it would seem to fit perfectly
> > with what I think the VM TF note should say.
> 
> Agreed, with some minor corrections to your paraphrase.
> 1. "Subject headings" is somehow a restriction of scope of PubSubj recommendation, which
> is about "subjects" in the widest possible sense, not only those defined in vocabularies.
> But this restriction is valid in VM TF scope.
> 
> 2. Topic Maps is the original application context for PSI. But as the introduction of the
> quoted recommendation hopefully makes clear, it's not the only one.
> 
> 3. The "human language label" requirement is also a restriction of the PubSubj
> recommendation, which simply states that a subject indicator should be "human
> interpretable". Think about the specific shade of blue defined by the RGB code #021A81.
> This is barely a "human language label", but the color itself is pretty well defined by
> the "human readable" subject indicator http://mediagods.com/tools/rgb2hex.html?464,294

Fine.

> > The open issues, on the other hand, seem to shade off into
> > community-specific philosophy with regard to the nature of the
> > terms identified and of the relationships among terms.  They
> > reflect that "evolving diversity" of choices about which "good
> > practice" may for valid historical reasons be still unclear --
> > things like "# versus /", the descriptive attributes of terms,
> > and details on publishing related documentation and metadata.
> >
> > Again, for such issues of "evolving diversity", I think the
> > VM TF note should simply summarize and point to ongoing work.
> > The VM TF membership would be hopefully diverse enough that we
> > could among ourselves come up with a reasonably representative
> > set of relevant citations.
> 
> Agreed

We seem to be agreeing :)  Would you want to sign up?
Tom

-- 
Dr. Thomas Baker                        Thomas.Baker@izb.fraunhofer.de
Institutszentrum Schloss Birlinghoven         mobile +49-160-9664-2129
Fraunhofer-Gesellschaft                          work +49-30-8109-9027
53754 Sankt Augustin, Germany                    fax +49-2241-144-2352
Personal email: thbaker79@alumni.amherst.edu

Received on Thursday, 17 June 2004 07:06:42 UTC