W3C home > Mailing lists > Public > public-esw-thes@w3.org > October 2005

Subjects and perspectives in SKOS : the jack of all trades ...

From: Bernard Vatant <bernard.vatant@mondeca.com>
Date: Thu, 20 Oct 2005 17:08:21 +0200
To: <public-esw-thes@w3.org>
Cc: "Michel Biezunski" <mb@infoloom.com>
Message-ID: <GOEIKOOAMJONEFCANOKCGEEIHBAA.bernard.vatant@mondeca.com>

Hello all

Browsing all those very interesting ongoing threads about possible extensions of SKOS,
relations with OWL, types of notes, terms-as-concepts, relevancy to terminology, etc ...
keeps bringing me back to the notion of *perspective* as currently explored by Michel
Biezunski [1], which I'm currently trying to bring along with my own current ramblings

In the following, the *highlighted words* are used according to Biezunski's definition. Or
at least they try to. Michel is in cc and will correct wherever I can get it wrong.

According to Biezunski's terminology, a skos:Concept is a *proxy* for some *subject*, as
any URI used in RDF is. The subject expressed by this proxy is in SKOS some abstract
concept, likely to be expressed otherwise in many specific formal or unformal ways, in so
many different schemes (thesaurus, taxonomy, ontology, terminology, ...) using so many
different languages (SKOS, OWL, UML ...) and matching representation rules, and those
expressions used in so many ways, for so many different purposes, in so many different
contexts. A combination of all of those defines a *perspective* on the subject/concept.

It's still unclear to me up to where a perspective on a skos:Concept can extend, were it
to be defined. It could include at least the rdf:Description, and/or all related
skos:Concepts in the same skos:ConceptScheme, or go as far as including this complete
scheme, and this is certainly not the end of the story, since a useful perspective should
certainly also include the purpose, ways, rules and context of use.

In any case, this opens different interesting questions.

The same URI can be used in different skos:Concept descriptions. So it has to be clarified
if the proxy for the concept is the URI or one of its rdf:Description.

The same skos:Concept can belong to, or be used in, a variety of perspectives. Not only
because it can belong to various skos:ConceptScheme(s), but because each of those schemes
can be used in different contexts, for different purposes, and in different ways :
indexing and classification (which seems to be SKOS primary purpose), but also text mining
and knowledge extraction, support for translation and publication tools ...

Among all possible properties of a skos:Concept, some are only relevant to certain
perspectives. Take for example the various kinds of notes, or properties on labels, or
lexical properties of terms ...

What does that lead us to? Interest for SKOS has attracted a variety of users with
different perspectives (and that is really really good), each of them pushing gently (only
gentle(wo)men here so far, very much appreciated) to allow the language to express, inside
the same description of a single skos:Concept any other property relevant to their
respective perspectives, at the risk of making at the end of the day such a description,
as Stella rightly pointed, the jack of all trades and the master of none.

Practically speaking, that means we are certainly at a point where SKOS should
- either "close its scope", by specifying as much as possible in which kind of
perspectives a skos:Concept is supposed to be used, and stick to the properties relevant
to such perspectives.
- or provide a way to express various perspectives, their respective context, purpose,
rules, and the way to "hub" them (this is where hubjects could be relevant).

The latter option if of course my favourite, even if much less obvious, it's certainly a
winner in the long run.

Enough for today. If there is some interest expressed in that, I can come up with more
formal ideas about it.



[2] http://www.google.com/search?q=hubject

Bernard Vatant
Mondeca Knowledge Engineering
(+33) 0871 488 459

Received on Thursday, 20 October 2005 15:08:59 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:45:23 UTC