RE: [VM] Draft of 2004-10-25

Thomas

Thanks for this clean draft, and for the clear definition of tasks.
I will be back about Published Subjects in a separate thread.

For now I've a few remarks, a specific one and more general ones :

The specific remark is about section 3.3.
I wonder what subtle difference you figure between the following tasks ...

        TASK: Bernard - Reuse of existing terms in a local context
        TASK: Everyone - Using terms outside of their original contexts

... and why I'm particularly challenged for the first one :))

The more general remarks are about the implicit or explicit use in the document of the
concept of "identification". I prefer this word, standing for a process, to the absolute
but elusive notion of "identity" - as is explained at length in recent posts at [2].

If one looks at the three first entries in the glossary, which should be the core of our
interagreement in spite of, or thanks to, their "deliberate fuzziness" ...

1.	Term                A named concept.
2.	Vocabulary          A set of terms.
3.	URI Reference       A globally unique identifier.

A fundamental question appears here that the document should address clearly (if not
answer) because other ones (including the bleeding edge ones) depend on the answer :

-- There are "things" (resources, subjects, entities ...) that we want to identify.

-- In the Semantic Web framework, things are identified by URIs

-- The things we are interested in are Terms in Vocabularies

-- A Term is a named concept.

I suppose that means that when we use a URI to identify a Term in a Vocabulary, we want to
identify a *concept* and not its *name*. Otherwise we would have written that a Term is a
*name for a concept*, and that we use the URI to identify the *name*, not the *concept*.
This is the classical debate between *terminological* vs *conceptual* views of Vocabulary.
I don't know if we want to stand clearly on one side of this debate. Some of the
vocabularies or languages we are about seem to stand clearly on the conceptual side of the
line. Seems that SKOS does, as Topic Maps Published Subjects do. I've never really been
sure about RDF and OWL being so definitive about it. In any case, the document should say
clearly at some point either if it stands on one side, and which, and maybe why that one,
or if it keeps agnostic, and in this latter case deal with both interpretations. This is a
very difficult question, but we can't sweep it under the carpet.

Coming back to 3.3, I'm very uneasy with the use of "meaning"(here as anywhere else) and
it seems that the wording of the issue in fact goes around identification without naming
it. The question of reuse is practical. If one reuses a term out of its original context,
to what extent can/should/may semantic applications (including human end-user brains) use
it for identification process the same way they would have used the original one? So the
question of "meaning" boils down to "Is it really the same concept, even if it bears the
same name and URI?". In this question, the key word is not "mean", it's "same".

More to come ...

**********************************************************************************

Bernard Vatant
Senior Consultant
Knowledge Engineering
bernard.vatant@mondeca.com

[1] "Making Sense of Content" :  http://www.mondeca.com
[2] "Everything is a Subject" :  http://universimmedia.blogspot.com

**********************************************************************************

Received on Thursday, 28 October 2004 22:30:48 UTC