Re: [SKOS] ISSUE-39A Grouping for Mapping? (was RE: [SKOS] A new proposal for ISSUE-39 ConceptualMappingLinks)

Hi Alistair,

I would really like to delegate to other issues this aspect of 
grouping/coordination in mapping. It would be better for ISSUE-39 to 
focus on mapping *links* and not to the question of what is mapped 
(apart from the point of view of domain/range assignments)

[Now for the content of your mail.]
When I browse some use cases of mapping, e.g. [4,5,6] (sorry, no time to 
even dream about a short use-case description for them :-( ) I find that 
kind of groupings quite some times.
For instance [4] has this:
hacker ^+ Computers + Crime , where "^" stands for "association" link 
(similar to related in my view) and "+" is of course this mysterious 
grouping.

My whole point here is that this "+" (or "AND" in other contexts) reads 
like a USE link
 From a document related to [2] I've got another example which is 
explicitly says that the use of AND should be interpreted as a 
combination at the indexing level.
In other terms, the semantic of AND would be: "if you have cs1:A 
exactMatch (cs2:B AND cs2:C) then you should consider all resources 
described by A as described by B and C at a same time". Which is very 
much related to post-coordination, and n-ary links between descriptors 
and non-descriptors, ISSUE-40 [8]!

In this respect I therefore think that your interpretation in [3] is a 
proper one.

And this convinces me further that we should really transfer this issue 
to ISSUE-45 [10] (on concept coordination) and ISSUE-40. And I wouldn't 
rule out the problem from the SKOS recommendation at this point, because 
it seems a number of people want to use these things. The resolutions on 
issues 40 and 45 should give a natural answer to this problem. So as 
long as there is no resolution for them, my position is: no resolution 
about coordinations in mappings.

Notice that in the end, among the grouping constructs AND, OR, NOT, I've 
seen only AND used (or refered to). To the best of my knowledge only the 
AIMS project [7] has used it.
which more or less fits what you said in [9] about ISO2788 and BS8723 
which mention AND but do not encourage the use of OR.

Cheers,

Antoine [interpreting a lot of different things now, I truly apologize 
if I misinterpreted some of them]

[1] <http://www.w3.org/2004/02/skos/mapping/spec/2004-11-11.html>
[2] <http://www.w3c.rl.ac.uk/SWAD/deliverables/8.4.html>
[3] <http://purl.org/net/retrieval>

[4] http://www.gesis.org/Forschung/Informationstechnologie/komohe.htm, 
also http://www.vascoda.de/
[5] http://macs.cenl.org
[6] http://www.d-nb.de/wir/projekte/crisscross.htm
[7] http://www.w3.org/TR/skos-ucr/#UC-Aims
[8] http://www.w3.org/2006/07/SWD/track/issues/45
[9] http://www.w3.org/2006/07/SWD/wiki/SkosDesign/ThesaurusPatterns
[10] http://www.w3.org/2006/07/SWD/track/issues/40

> Hi all,
>
>   
>> (ISSUE-39A) Should "grouping" constructs for mapping be 
>> included, and if so, what are their semantics?
>>     
>
> My position: I do not think it's a good idea to include any "grouping" constructs such as "AND", "OR", "NOT" in the SKOS recommendation, because it would require substantial additions to the theoretical foundation of SKOS.
>
> The SKOS Mapping vocabulary was originally defined in two documents, [1] and [2]. Both of these include "AND", "OR" and "NOT" constructs. There are some conflicts between [1] and [2], but both share the basic idea that "AND", "OR" and "NOT" are "groups" of concepts. 
>
> How are these to be interpreted? Again, there are some conflicts between [1] and [2]. [1] talks about set operations (intersection, union, complement, see [1]#AND, [1]#OR, [1]#NOT). [2] talks about set operations too (see [2]#3.a), however [2] also talks about "intension", e.g. "Use the AND construct to create a mapping target that represents the common intension of two or more concepts."
>
> There are difficulties with the set-based interpretation of "AND", "OR", and "NOT". 
>
> Firstly, to give a proper account of this interpretation, you need a notion of a "query" which can be evaluated with respect to an "index". 
>
> If you have both of these, mapping statements involving "AND", "OR" and "NOT" constructs can make sense as mappings between *query expressions*. Mappings between composite query expressions can be given a purely operational interpretation, which defines procedures for translating either indexes or query expressions. See [3], chapter 7.
>
> This is the only way I know of providing a solid foundation for "AND", "OR" and "NOT" expressions. To introduce this foundation for SKOS, we would first need to establish a notion of a "query", which is no small task.
>
> For these reasons, at the moment I would vote against including "AND", "OR" or "NOT" grouping constructs in the SKOS recommendation.
>
> However, I wouldn't rule out a notion of "coordination". I think "coordination" can be given an informal explanation, for which there is some precedent.
>
> Anyway, that's all I have for now, cheers,
>
> Alistair.
>
> [1] <http://www.w3.org/2004/02/skos/mapping/spec/2004-11-11.html>
> [2] <http://www.w3c.rl.ac.uk/SWAD/deliverables/8.4.html>
> [3] <http://purl.org/net/retrieval>
>
> --
> 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 Wednesday, 5 December 2007 20:31:30 UTC