Re: notes at contepts vs notes at terms

In a posting way back on 1 March 2005, I mentioned a couple of problems 
with the underlying SKOS model in terms of thesaurus management. The one 
issue seems to have attracted a lot of attention ('notes at term level') 
and some progress seems to be being made on accommodating that issue. 
However I feel I also have to raise again the second problem, that of 
semantic associations between non-preferred terms (or tokens as Mark 
suggests).

Frequently in multilingual thesauri, we want to capture a relationship of 
equivalence between different tokens used as altLabels in different 
languages. This feature is frequently used in multilingual thesauri such as 
the OECD Macrothesaurus, the ILO Thesaurus or Agrivoc (there are 
undoubtedly others, but these are the ones that I know about). In semantic 
terms, these altLabel tokens have some close relationship (usually they all 
equivalently express an underlying shared concept [1] that has been 
specifically excluded from the accepted concepts within the thesaurus 
itself, often for reasons of literary warrant). In practical terms, the 
different altLabel tokens are then managed together as a single unit: for 
example, in doing a translation into a new thesaurus language, there would 
be a special effort to try to find an equivalent altLabel in the target 
language corresponding to that excluded concept, and in revising or 
adapting the thesaurus to local conditions and use, the non-descriptors may 
be promoted together to descriptor level (or descriptors demoted to 
non-descriptors, while retaining the semantic relationship between the 
altLabel tokens that exists in the original descriptor/concept record).

If SKOS is meant to be a way of interoperating for query expansion and 
similar kinds of uses, this feature is not (I think) necessary. In fact, in 
my personal opinion, I think at this stage we should be keeping SKOS as 
simple as possible to promote wide implementation.

However if we talk about SKOS as a way of exchanging thesaurus information 
for purposes of thesaurus management (which I know is one of the uses that 
Alistair talks about), then it is both a necessary and reasonable 
requirement. Otherwise multilingual thesaurus won't be able to use SKOS for 
management purposes without a significant loss of information.

I'm hoping that some of the proposals being put forward for accommodating 
notes on altLabel tokens will also be able to accommodate this multilingual 
requirement.

Ron
[1] Sorry if this use of the word concept offends but I don't know what 
other word to use here.

At 12:29 1/11/2005, Mark van Assem wrote:
>Hi Alistair,
>
>I'm leaving the "use cases" that I wrote about for later, in this post I 
>to summarize what I think Sue Ellen Wright, Bernard Vatant, Phil Carlisle 
>and Stella Dextre Clark have been saying (please correct me if I'm wrong!) 
>[1,2,3,4,5]. I'm hoping Sue and Stella can find time to provide some more 
>examples.
>
>About the word "term": I agree that it is overloaded. Should I call "the 
>thing in the range of the skos:prefLabel and skos:altLabel properties" a 
>Label or a Token?
>
>If a class e.g. Label is introduced instead of the literal currently 
>defined as the range of skos:prefLabel and skos:altLabel, additional 
>information can be attached to Labels. The categories of information that 
>can be attached to instances of a class Label or Token are (summarizing 
>other people's posts):
>
>- scope notes for terms, also referring to other terms to use [4]
>- lexical information about the term [2,5]
>- scope of usage of the lexical term [5]
>- etymological, register-related, standardization
>related [2] (I hope Sue can find time to clarify this further)
>
>(what follows is not summarizing [1-5])
>
>Furthermore, some examples from MeSH [6]:
>
>- TermUI (local identifier)
>- date created
>- source thesaurus (MeSH groups different thesauri into one)
>- abbreviation
>
>which mostly fall under a category "editorial information". I also 
>remember someone posting that some thesauri attach different definitions 
>to different terms.
>
>A class Label would make it possible to extend the SKOS schema for 
>categories of information (attached to Label) we can't foresee right now.
>
>You provide an alternative way to attach notes to labels [7]:
>
>ex:conceptA a skos:Concept;
>    skos:prefLabel 'Animals';
>    skos:altLabel 'Fauna';
>    skos:editorialNote [
>      skos:onLbl 'Fauna';
>      rdf:value 'Check with Mr.X. whether to keep "Fauna".';
>    ];
>
>I think this is a solution, but in principle this method could be used 
>everywhere you normally use a class to group information about an entity. 
>I think the more usual way to do this in RDF or OWL is introduce a class. 
>Also, this is harder to maintain (changing skos:altLabel 'Fauna' without 
>changing skos:onLbl 'Fauna' leads to errors).
>
>Of course we can choose not to support this kind of information attached 
>to terms, but then we should say so explicitly.
>
>Cheers,
>Mark.
>
>
>[1] http://lists.w3.org/Archives/Public/public-esw-thes/2005Oct/0119
>[2] http://lists.w3.org/Archives/Public/public-esw-thes/2005Oct/0109.html
>[3]http://lists.w3.org/Archives/Public/public-esw-thes/2005Oct/0101.html
>[4] http://lists.w3.org/Archives/Public/public-esw-thes/2005Aug/0019
>[5]http://lists.w3.org/Archives/Public/public-esw-thes/2005Aug/0018
>[6] http://www.nlm.nih.gov/mesh/xml2006sample.txt
>[7]http://lists.w3.org/Archives/Public/public-esw-thes/2005Aug/0017.html
>
>
>--
>  Mark F.J. van Assem - Vrije Universiteit Amsterdam
>        mark@cs.vu.nl - http://www.cs.vu.nl/~mark
>
>