re: model-model mapping

> Could anyone please point me in the direction of any work done on ways of
> mapping between different vocabularies, schemas and their models.

The logical notion of theory morphism or interpretation (between
theories) may be useful here.  The key idea is to translate from one
language to the other in a way that preserves meaning, specifically,
theorems are preserved under translation.  The translation is
typically specified by a symbol-to-term map.

At Kestrel Institute we use interpretations to refine one
specification to another.  Any data structure in the source spec can
be translated into a data structure in the target spec, and then
operated upon by algorithms in the target.  We are pursuing the use of
interpretations as a basis for ontology translation in the DARPA DAML
project.

Many texts on logic cover interpretations between theories
(e.g. Schoenfield, Enderton).  Originally, they were developed as a
way to study the relative consistency of two theories.

Cheers,
 Doug

> -------- Original Message --------
> Subject: model-model mapping
> Resent-Date: Sat, 12 Jan 2002 17:14:34 -0500 (EST)
> Resent-From: www-rdf-logic@w3.org
> Date: Sat, 12 Jan 2002 23:10:28 +0100
> From: "Danny Ayers" <danny666@virgilio.it>
> To: "Www-Rdf-Logic" <www-rdf-logic@w3.org>
> 
> Could anyone please point me in the direction of any work done on ways of
> mapping between different vocabularies, schemas and their models.  There's
> obviously been a fair bit of work been done on things like mapping RDBMS schema
> to XML schema or object models, but what I'm after is a more generic, 'meta'
> form of this. Terms like 'equivalentTo' would work in this context, but what
> would be really nice would be to find a whole RDF Schema in this space ;-)
> 
> The particular application I have in mind only needs a narrow aspect of this :
> more or less one-to-one mapping between an arbitrary domain-specific graph based
> model (such as that of a hierarchical organisation) and a general-case graph
> (just nodes & arcs here), but the ability to map from one model to another
> enables the reasoning facilities (i.e. algorithms) of the second model to be
> applied to the first, which is an approach that strikes me as being ideal for SW
> work, but rather neglected (ok, this is something that is done a lot through the
> back door, e.g. using the Rete algorithm in Jess plugged into from Protege, but
> the mapping is, meta-speaking, usually hard-coded). If I do have to come up with
> my own RDFS, then it might as well be based on a reasonably general case (rather
> than the feeble attempts I've tried so far).
> 
> Cheers,
> Danny.

Received on Monday, 14 January 2002 18:35:09 UTC