W3C home > Mailing lists > Public > public-esw-thes@w3.org > June 2007

Re: [SKOS] ISSUE-33 "SimpleExtension" proposal (was RE: [SKOS] "Lexicalization" or "term"?)

From: Antoine Isaac <aisaac@few.vu.nl>
Date: Fri, 15 Jun 2007 17:14:47 +0200
Message-ID: <4672ACE7.3010600@few.vu.nl>
To: "Miles, AJ \(Alistair\)" <A.J.Miles@rl.ac.uk>
CC: SWD WG <public-swd-wg@w3.org>, public-esw-thes@w3.org

Hello Alistair,

> Hi Antoine,
>
> I just took a look at the latest revision of the "SimpleExtension" proposal [1] for ISSUE-33. Interesting :)
>   
Well I didn't write this proposal, just typed it for Guus for whom the 
wiki interface is especially difficult to deal with. But I'm interested 
in it so I'll give my answer ;-)
> Under this proposal, is there a bijection [2] between the extension of skos:Label and the set of RDF plain literals?
>
> In other words, is there only one skos:Label for every plain literal, and vice versa?
>
> Or, under this proposal, is there only a surjection [3] between the extension of skos:Label and the set of RDF plain literals
>
> In other words, is there only one plain literal for every skos:Label, but one or more skos:Label for every plain literal?
>   
> This is absolutely crucial to exploring the consequences of the proposal. It is crucial because it bears on the conditions under which it makes practical and logical sense to assert the identity of two individuals of type skos:Label. This is the fundamental question that all proposals following the "terms-as-classes" pattern must address.
>
> Cheers,
>
> Al.


One first thing: if you refuse the entailment rules in section 6 [1], 
then there is of course no link between the two sets of entities ;-) 
Don't forget Guus has proposed them for discussion. Which we are doing 
now, nice.

Now if we accept them, which I'm in favor of for the moment, there is a 
link, of course. But this is only surjective: all skos:Label can be 
associated to a literal, but not every literal can be associated to a 
skos:Label. It is likely that "nferereg987fwef" will never be used to 
label a concept. The extension of skos:Label, assuming the entailment 
rules in section 6 of [1] is indeed in bijection with these literals 
that are used to label some skos:Concept.


Now for the identity conditions for these skos:Labels. In a first 
analysis I think two skos:Label should be equal if their (unique) 
rdfs:label are equal. This includes of course the language attribute 
value. Which is equivalent to the literal identity condition in RDF I think.


Notice that we can also discuss this problem related to the mail of Jon 
on context-dependant application [4], there is the problem of the 
dependance of the links we would like to assert between these labels 
with respect to specific contexts. E.g. in some situations "goods"@en is 
an antonym of "bads"@en, while in other contexts (when goods are 
manufactured products) it is not.
A solution would be to consider the skos:Labels as context-dependent, 
e.g., practically, by linking them to a ConceptScheme. But this would 
amount to have them lose their neutrality, becoming these terminological 
entities (with some meaning attached to it) that I wanted to avoid (see 
other mails in this thread).

So this is why I would prefer to have the context dependency rather 
attached to lexical links statements (good antonym bad) and keep for 
labels neutral lexical entities which are in bijection with literals, 
that is language-situated ordered sets of characters (sorry, end of the 
week appraoching ;-).

Cheers,

Antoine

[1] 
http://www.w3.org/2006/07/SWD/wiki/SkosDesign/RelationshipsBetweenLabels/ProposalThree
[4] http://lists.w3.org/Archives/Public/public-swd-wg/2007Feb/0187.html

PS: was the end of your mail a typo?
>
>
> [2] <http://en.wikipedia.org/wiki/Bijection>
> [3] <http://en.wikipedia.org/wiki/Surjective_function>
>
>
>
> This proposal makes me happy, not because I support it necessarily, but because , because there is enough information there to understand the intended semantics There are two fundamental questions this propsal
>
>
> It seems to me that [1] intendes the skos:Label class is equivalent to the class of RDF
>
> [1] <http://www.w3.org/2006/07/SWD/wiki/SkosDesign/RelationshipsBetweenLabels/ProposalThree?action=recall&rev=14>
>
> --
> Alistair Miles
> Research Associate
> Science and Technology Facilities Council
> Rutherford Appleton Laboratory
> Harwell Science and Innovation Campus
> Didcot
> Oxfordshire OX11 0QX
> United Kingdom
> Web: http://purl.org/net/aliman
> Email: a.j.miles@rl.ac.uk
> Tel: +44 (0)1235 445440  
>
>   
Received on Friday, 15 June 2007 15:14:53 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:58 GMT