RE : SKOS Mapping Vocabulary / http://www.w3.org/2004/02/skos/mapping/spec/

Dear Martin,

These comments are very interesting. However, I have questions on them, of course ;-)

1. Do you have more precise comments on places where the SKOS mapping document is unclear regarding the use of mapping relations for SKOS concepts? I think it it not the ambition of SKOS mapping to give an mapping language for ontologies in general. Of course people developing SKOS concept schemes might decide to interprete some their skos:Concepts as OWL classes, and that might cause SKOS mapping links to hold between OWL classes, but I guess this is a side effect, not something SKOS really advices.
So if you have specific sentences that lead to ambiguity, please send them, they will help the work of whoever edits a next version of this SKOS mapping

2. I think you are really pessimistic about interpretations in classifications. Indeed the example you give about a same class being applied to described products or persons seems a very weird one. I will read your paper, but I have indeed a counter example in the place I work, where they have to describe 'children books' and 'books about children books'. And they have introduced two concepts, 'children book' and 'children bookS' to avoid the confusion.
I would be anyway really interested in getting feedback of people on this lists. For me a classification, even a semantically fuzzy one, has at least some coherence with respect to the objects that should be described with it (books, websites, products). Perhaps a situation such as the one you describe could happen in different systems, that ignore the rationale of the classification design? Then clearly something would have gone wrong...

Cheers,

Antoine

-------- Message d'origine--------
De: public-esw-thes-request@w3.org de la part de Martin Hepp
Date: mar. 05/06/2007 10:39
À: danbri@danbri.org; public-esw-thes@w3.org
Objet : SKOS Mapping Vocabulary / http://www.w3.org/2004/02/skos/mapping/spec/
 

Dear Dan, all:

I understand that the SKOS Mapping Vocabulary

http://www.w3.org/2004/02/skos/mapping/spec/

aims at providing a lightweight, consensual formalism for expressing
mapping relations

a) between elements in the same SKOS vocabulary,
b) between elements in multiple SKOS vocabularies, and
c) between elements in a SKOS vocabulary and elements in an RDF-S or OWL
ontology (I am not completely sure whether that is intended but can also
not read that from the Web page).

This is of course a very important contribution. However, I would like
to point out that there is one important difference between
"categorization systems" (or "classifications") on one hand and
ontologies on the other: Classifications *do often NOT have a unique,
context-independent notion of the meaning of a class* - rather, the
intended semantics of a category may vary by context, and people
exchanging classified data must implicitly or explicity agree upon the
context of usage.

For example, we can use the same hierarchical classification of products
and services for classifying

a) actual products and services,
b) staff member who have expertise in particular types of products and
services, or
c) expenses related to a particular type of products and services.

Now, the overall semantics of a category is then no more specific than
"anything that can in any reasonable context be subsumed under the label
XYZ". The problematic thing is, however, that people using
classifications assume a much more restricted intended meaning per category.

Thus, it is problematic to assume that SKOS concepts are also ontology
concepts - a notion that may, if not explicitly intended by SKOS, at
least be assumed by many users of SKOS, and seems to pervade the Mapping
Vocabulary Spec.

We have discussed this issue and how categories in classifications can
be turned into ontology classes in [1]. In our opinion, deriving
concepts with a unique, context-independent semantics from categories in
classifications includes some important modeling choices, which affect
the usefulness of the resulting ontology.

Best

Martin

[1] Martin Hepp, Jos de Bruijn: GenTax: A Generic Methodology for
Deriving OWL and RDF-S Ontologies from Hierarchical Classifications,
Thesauri, and Inconsistent Taxonomies, Proceedings of the 4th European
Semantic Web Conference (ESWC 2007), June 3-7, Innsbruck, Austria, in:
E. Fraconi, M. Kifer, and W. May (Eds.): ESWC 2007, LNCS 4519,  Springer
2007, pp. 129-144.
PDF at http://www.heppnetz.de/files/hepp-de-bruijn-ESWC2007-gentax-CRC.pdf