SUO: Re: identity of terms when merging ontologies
Seth,
Theoretically, I discussed type identity (synonymy) in the context of
ontology sharing in the ISKO'6 paper
[http://www.ontologos.org/Papers/ISKO6/ISKO6.pdf]. From the formal
standpoint, the IFF will include this capability in the quotienting aspect
of colimits. As I mentioned before
[http://suo.ieee.org/email/msg06222.html], in both of the mathematical
contexts of classifications and concept lattices, we will be able to specify
and construct colimits. We will also be able to do this in the context of
IFF models. But the model theory axiomatization will not appear until IFF
version 3.0.
Robert E. Kent
rekent@ontologos.org
----- Original Message -----
From: "Seth Russell" <seth@robustai.net>
To: "Adam Pease" <apease@ks.teknowledge.com>; "John F. Sowa"
<sowa@bestweb.net>
Cc: "swag-dev" <swag-dev@yahoogroups.com>; "SUO"
<standard-upper-ontology@ieee.org>
Sent: Saturday, September 01, 2001 11:11 AM
Subject: SUO: identity of terms when merging ontologies
>
> From: "John F. Sowa" <sowa@bestweb.net>
>
> > Given the axioms for theory A and the axioms for theory B,
> > you can create the axioms for the combined theory C
> > by using the cat (concatenate) command in Unix:
> > cat A B > C
> > This creates a new file C, which contains all the axioms for
> > both A and B.
>
> Well in relational db terms: (a) create a data base containing both sets
of
> records , (b) add a field that specifies the original source of the
axiom,
> and (c) adjust each term of each statement to use the same namespace
> strategy.
>
> > There are now two more questions one would like to ask:
> > 1. Is the new theory consistent?
> > 2. Are any of the axioms redundant?
>
> I think we need to add a step preceding those two. The identity of the
> terms used in both theories need to be matched. Of course if the theories
> were originally designed with the same terms, this step is trivial, but
> otherwise it could approach a can of worms. For example: Doug likes
'isa'
> , Adam likes 'instance', and DAML likes 'type'. Exclusive of the
> difference in the axioms which, I assume, will be dealt with in your steps
> (1&2 above) these are pretty much the same concept and, I would think, any
> ~merged~ ontology would need to ~smush~ [1] them together somehow. But
if
> we just concat the files after making adjustment (c) above, they just end
up
> being in different namespaces.
>
> How do we resolve this identity problem?
>
> [1] http://rdfweb.org/2001/01/design/smush.html
> [2] http://lists.w3.org/Archives/Public/www-rdf-interest/2001Jan/0030.html
>
> I cc'd SWAG [3] on this because it was created to deal with problems like
> this.
>
> [3] http://purl.org/SWAG/
>
> Seth Russell
>
>
>