RE : SCOS?

Hi Sue Ellen, Aida et al,

Thanks for the discussion. It won't be a surprise that I favor keeping to SKOS myself ;-)

A quick question to Sue Ellen, though: are you really using OWL to directly represent your terminologies? Or to put it differently, are you creating terms as direct instances of owl:Classes, or as instances of your own classes (e.g. mymodel:Term)
Because in a way, SKOS uses OWL (the SKOS model is defined as an OWL ontology). And I would expect other (terminology) models to be positioned at the same epistemological level, so I'm a bit surprised that you constrast your approach and the SKOS one so strongly...

Antoine


-------- Message d'origine--------
De: public-esw-thes-request@w3.org de la part de Sue Ellen Wright
Date: sam. 01/11/2008 02:13
À: Aida Slavic
Cc: Skos
Objet : Re: SCOS?
 
Hi, All,
At the risk of too many layers ...

On Fri, Oct 31, 2008 at 3:27 PM, Aida Slavic <aida@acorweb.net> wrote:

>
> Sue,
>
> Thanks for your comment - it is well worth pondering about. We only need to
> know how little we know about cognitive processes (and cognitive linguistics
> and semantics) to be very careful about SCOS/SKOS issue.
>
> Maybe SKOS may serve theoretically for : a)language-purposed vocabulary b)
> subject indexing vocabulary and even c) store 'document vocabulary' (i.e.
> natural language of the document itself) d) real world representation


... Actually, we're finding that we need to go to OWL DL for a lot of what
we do in language-purposed  terminology because we need some extra
functionalities, but by the same token, I haven't had time to go back and
look at some the "new stuff" to see if we could go back. I'm afraid I'm sort
of enamored of DL at this point! Our colleagues in lexicography (which is
word centered and more complex) are using Full Owl + rule constraints.

>
> But I have many doubts that plunging into this would help.
> SKOS' started with the easiest task: the one of supporting subject indexing
> languages: a little formalized sets of well organized vocabularies. This is
> where we all know it can make a difference and bring many benefits very
> quickly and with least effort. But even here it seems awfully hard to agree
> about basics
>

And this is where SKOS is strong. I'm not convinced that it should be
stretched because as noted, why add stuff to replicate OWL DL or other
solutions if those solutions are working just fine for other areas of
activity. I didn't go into all of this in my email, but maybe it needs to be
said. I'm very much inclined to think that the complexities of concept
modeling on the epistemological may be closer to some of these other
solutions than they are to subject-purposed vocabularies.

> that I call Knowledge Representation Resources. Howard has just
> broadened that view. Some in this group will have heard me invoke the
> distinction by Svenonius and Doug Tudhope between subject-purposed
> vocabulary and language-purposed vocabulary -- are we talking about
> /resources about butterflies /(in which case our instances are
> books/articles/webpages/etc. that have butterflies as their object of
> study), or are we talking about /butterflies/ (mostly beautiful,
> graceful insects that feed on flower nectar, etc.). Or in Howard's case,

And what an important distinction this is, indeed.
> KOS such as thesauri, documentary classification and subject heading
> systems are created for mediation of knowledge. They are 'interpreters' of
> recorded knowledge and external to documents as carriers of recorded
> knowledge.
>
> They have three sides to them:
>
> 1) They "mediate"  and control relationships between thoughts (concepts)
> and language - so at first glance they appear to be the same as any other
> language
> But their function and strength is precisely in their being
> detached/separate from the language of the document itself and from the
> language user would use when searching.


Absolutely. I couldn't agree more.

>
> 2) they are formalised systems based on certain convention (scientific,
> educational, cultural...) in order to achieve predictability. They interpret
> language terms within a certain knowledge context: certain subject,
> scientific point of view, task, audience or disciplines, knowledge field
> etc. Hence the 'same' concept in one KOS will have different
> broader/narrower relationships in different KOS...
>
> 3) they may also contain 'vocabulary aparatus' to express relationship
> between the external (objective?! no Popper intended) knowledge and the one
> recorded in the document, or its presentation in a given document  - as well
> as relationship between knowledge and document as a carrier
>
> So for people using SKOS would best serve its purpose of supporting this
> kind of resource discovery by allowing us to manage links between: KOS >
> document-metadata > document itself,  independently.
> For the first phase of SKOS some of these things above were not relevant -
> but when OWL got involved then the purpose/task, function and
> information/metadata architecture became very important.
> And somehow I don't have the feeling that there is a clear understanding of
> the above.
>
> It is going to be very interesting to see how discussion on language
> vocabulary (morphology, syntax, grammar) would go should we decide to follow
> SCOS part :-)


We presented a paper on our decision to use OWL DL for language-purposed
terminology at LREC, which prompted someone to ask me "to please send my
criticism of SKOS" -- to which I responded that I have none to speak of,
that I think SKOS is highly useful and powerful for the precise purpose for
which it is being designed. My choice of OWL DL has everything to do with
what I have concluded is a different purpose and doesn't reflect any
inherent weakness in SKOS. The confusion arises in the fact that we all
share certain terms and ways of addressing issues, but in the different
approaches, those terms, their underlying concepts, and the framework in
which they function reside on different conceptual and methodological
planes.

Best regards
Sue Ellen

>
>
> Aida
>
>
>


-- 
Sue Ellen Wright
Institute for Applied Linguistics
Kent State University
Kent OH 44242 USA
sellenwright@gmail.com

Terminology management: There is unfortunately no cure for terminology; you
can only hope to manage it. (Kelly Washbourne)