W3C home > Mailing lists > Public > public-swd-wg@w3.org > February 2007

Re: [SKOS] relationships between labels

From: Antoine Isaac <aisaac@few.vu.nl>
Date: Wed, 21 Feb 2007 09:23:45 +0100
Message-ID: <45DC0191.8050905@few.vu.nl>
To: Alistair Miles <a.j.miles@rl.ac.uk>
CC: SWD WG <public-swd-wg@w3.org>

Hello Alistair,

Everything amounts to choosing a name for the class representing 
terms/label/whatever that would not have too much apriori 
interpretation. That's why I didn't make a choice, hoping this was 
enough to raise the problem (notice that Label is not perfect either, 
since it collides with the property rdfs:label).
Perhaps we could for the moment try something like "LabellingEntity", 
then. Each labelling entity would be required then to have one (unique) 
"preferred lexical value" (the equivalent of the prefLabel and altLabel, 
depending if we were at a preferred LabellingEntity or an alternative 
one), and possible some alternative forms (but here we don't care). This 
could be compatible with [2], I think.

Also I had already read your [3]. And agreed with many things there. The 
only objection I would raise is that, as designers of SKOS, this WG has 
some freedom decide for the interpretation of the constructs we 
introduce (provided of course it is not too counter-intuitive). Just 
think of skos:Concept. SKOS contains a highly debatable entity, assigns 
it a highly ambiguous term, and restricts its interpretation in a very 
specific way, which could be argued against by many (just think of 
skos:Concept vs. owl:Concept). And yet there is interest for SKOS.



PS: this discussion aprat I will not fight till death for having 
"Term-as-class" in SKOS. But I would like any refusal to be further 
discussed. On the SKOS list there were arguments against it, arguments 
for it (I have tried to note them, I'll push them to the wiki soon, I 
hope), but the final decision was made in a way that I could not really 
keep track of, at least browsing the mailing list...

> Hi all,
> Just a quick note to add to the record that the fundamental difficulty 
> with the "Term-as-Class" approach is agreeing on the semantics, and 
> especially the identity conditions, for the members of the suggested 
> new class.
> In my experience, while some people may appear to hold superficially 
> similar notions of what a "Term" is, when you try to express those 
> notions formally you can arive at some interesting contradictions.
> For example, in [1] Antoine essentially states that a "Term" cannot 
> have more than one lexical value. However, others would disagree with 
> this condition, and would consider e.g. "organisation" and 
> "organization" to be "variant forms" of the same "Term" (this is the 
> FAO/AGROVOC approach, see [2]). I don't mean to suggest that either is 
> "correct", I just want to demonstrate that these two points of view 
> are actually irreconcilable.
> A related problem is, if two "Terms" have the same lexical value, are 
> they therefore the same "Term"? Or can two "Terms" have the same 
> lexical values, but still be different entities?
> It's worth noting also that, in the DCMI community, "Term" can mean 
> property or class or concept. i.e. nothing at all to do with lexical 
> values.
> A while ago I tried to write up a discussion of how different points 
> of view on what a "Term" is lead to drastically different (and 
> inconsistent) logical definitions - it's at [3], not very clear but 
> the examples might help.
> Cheers,
> Al.
> [1] 
> http://www.w3.org/2006/07/SWD/wiki/SkosDesign/RelationshipsBetweenLabels?action=recall&rev=6 
> [2] http://owl-workshop.man.ac.uk/acceptedPosition/submission_31.pdf
> [3] http://isegserv.itd.rl.ac.uk/blogs/alistair/archives/25
> Antoine Isaac wrote:
>> Hi all,
>> I have written on 
>> http://www.w3.org/2006/07/SWD/wiki/SkosDesign/RelationshipsBetweenLabels 
>> a first attempt regarding the action:
>>> [PENDING] ACTION: Antoine to capture the issue on capturing 
>>> relationships between labels [recorded in 
>>> [50]http://www.w3.org/2007/01/22-swd-minutes.html#action04]
>> Perhaps the most important, concerning the action itself, is the 
>> 'motivation' part that gives example of such links between labels.
>> The 'solutions' part goes into discussion on ways to solve the issue, 
>> inspired by the material found via [1]  and the page Alistair 
>> recently wrote about Annotation Patterns [2].
>> Any further attempt to go into a solution in terms of SKOS model 
>> shall be worked out in cooperation with the issue on annotations on 
>> labels, I think.
>> Cheers,
>> Antoine
>> [1] 
>> http://www.w3.org/2004/02/skos/core/proposals.html#thesaurusRepresentation-11 
>> [2] http://www.w3.org/2006/07/SWD/wiki/SkosDesign/AnnotationPatterns
Received on Wednesday, 21 February 2007 09:10:36 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:07:49 UTC