[SKOS] TR : ISSUE 47 MappingProvenanceInformation

Hi Margherita (I cc your mail to the SWD list),

(And sorry Guus I promise this will be my only interfering with the 
issue you've just seized from me ;-)

> -------- Message d'origine--------
>
> De: public-esw-thes-request@w3.org de la part de Sini, Margherita (KCEW)
> Date: mar. 05/02/2008 18:04
> À: public-esw-thes@w3.org
> Objet : ISSUE 47 MappingProvenanceInformation
>
>
> ISSUE just opened after today conference.
> I would mention is very important for us because based on different 
> needs we
> may have different mappings.
>

That's indeed a very important motivation for such a requirement.

>
> I propose to assign a creator or owner to the mapping so to idenfity the
> provenance. and again by reusing if possible something already 
> existing e.g.
> dc:author or dc:creator (forgot which one is).
>
>
> Regards
> Margherita


The problem is that the issue may refer to indvidual "mapping 
statements", e.g. [ex1:cat skos:exactMatch ex2:chat].
So applying your solution is technically feasible, but would require RDF 
reification. We are here in a situation very similar to ISSUE-36 
regarding containment of semantic relationships in concept schemes.
And since RDF reification is not popular, we cannot go for this solution.

Indeed, two solutions are possible:
1. Creating a kind of "mapping scheme", that could be treated as an RDF 
named graph. Knowing that a specific MappingScheme object has for 
instace ex:margherita as dc:creator and that it is the context in which 
the mapping [ex1:cat skos:exactMatch ex2:chat] was asserted, then you 
could by using appropriate SPARQL queries retrieve your provenance 
information. This is very similar to the solution we accepted for 
ISSUE-36 [1]

2. Creating a kind of "reification" for the mapping, similar to the 
pattern Alistair used for ISSUE-26 [2]
Instead of [ex1:cat skos:exactMatch ex2:chat] (or complementary to it) 
we would assert the following triple
_:b1 rdf:type MappingRelation;
  skos:mappedConcept1 ex1:cat;
  skos:mappedConcept2 ex2:chat;
  skos:mappingRelationType skos:exactMatch;
  dc:creator ex:margherita.

This is actually what is done in current ontology alignment community, 
e.g. the format used for the OAEI evaluation campaigns [3, 4-p5], which 
introduces mapping "cells". These cells are gathered in "alignments" 
using simple RDF statements. Conitnueing my fictional SKOS namespace 
(but everything can be represented using the vocabulary from [3])
ex:myMappingScheme rdf:type skos:MappingScheme;
  skos:includesMapping _:b1.

Notice that the two solutions have their strong and weak points:
- 1 is closer to the way SKOS paradigmatic relationships are expressed, 
but is less flexible in terms of representation: things will become 
messy if "mapping schemes" aggregate mappings from various origins
- 2 is more powerful at representing provenance information (you can 
distinguish between the creator of the "mapping scheme" and the creator 
of each mapping statement), but has clearly a technical flavor (far from 
the way SKOS models its semantic relationships)

Best,

Antoine

[1] http://www.w3.org/TR/skos-reference/#L9287
[2] http://www.w3.org/TR/skos-reference/#L2914
[3] http://oaei.inrialpes.fr/2007/
[4] http://gforge.inria.fr/docman/view.php/117/251/align.pdf

>
>
>
>
>

Received on Tuesday, 5 February 2008 19:44:07 UTC