Re: Lang: owl:ontolgy

From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: RE: Lang: owl:ontolgy
Date: Fri, 13 Sep 2002 14:14:47 +0200

> >
> > > DAML+OIL seems to assume that we can hack URIs to see that
> > >
> > > http://example.org/onto#Potato
> > >
> > > is defined in
> > >
> > > http://example.org/onto
> >
> > I would be interested in seeing how you come up with this conclusion, and
> > where in DAML+OIL it is used.
> 
> OK.
> 
> DAML+OIL appears to be based on the RDF graph, although it has a default
> form in XML.
> 
> A small ontology looks something like:
> 
> 
> <rdf:RDF>
>   <daml:Ontology rdf:about="" daml:version="1.0">
>      <rdfs:description>
>         An ontology about Potatos.
>      </rdfs:description>
>   </daml:Ontology>
>   <daml:Class rdf:ID="Potato" />
> </rdf:RDF>
> 
> How are we meant to know that the version number has anything to do with the
> Potato class?

We don't, and it doesn't matter in DAML+OIL, as nothing depends on it.

DAML+OIL has problems with ontologies and documents.   Part of DAML+OIL
acts as if an ontology is a document and part of DAML+OIL acts as if an
ontology is a resource.  There is no connection between things like classes
and ontologies as resources except through ontologies as documents.

> Particularly if we start with the class ...
> 
> (Suppose we have loaded three or four files in).
> 
> Well, we have to know which file we loaded "Potato" from, or rather where it
> was originally defined. Since we have no explicit information about this, a
> moderately robust way to proceed is to look at its URI and remove the
> fragment id "#Potato". This will get us down to the document URI, and lo and
> behold we can find a version number.

Well, I suppose you might be able to do this, assuming that you wanted to,
but I don't see anything in DAML+OIL that needs to do ti.

[...]

> To answer Peter's question.
> With this sort of example I can see a means for getting from Potato to
> version 1.0 by URI cracking. I don't feel confident that any other method
> would be robust in practice.

I agree that there is no good method for getting from a resource to an
ontology except by examining the URI of that resource.  However, I don't
see what doing this gets you in DAML+OIL.

Now there are certainly extra-logical reasons to want to get from a
resource, or axiom, or whatever to some indication of
where/when/why/whatever things were said about it.  DAML+OIL doesn't
provide much guidance or utility for this.

If you want to associate to ontologies, I think that it would be much
better to be able to have ontology documents look something like:

<swol:Ontology [extra-logical information about the document]>

  [axioms, etc, including extra-logical information]

</swol:Ontology>

but this isn't RDF.

> Jeremy

peter

Received on Friday, 13 September 2002 08:48:58 UTC