RE: DTTF: List Ontology test case

From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: RE: DTTF: List Ontology test case
Date: Thu, 2 May 2002 15:06:38 +0100

> 
> Peter (stating his false beliefs about Jeremy):
> > > > Well, I don't think that you get want you need out of this
> > sort of thing.
> 
> 
> Peter again:
> > So, for DAML+OIL lists to behave as expected, there needs to be a
> > *syntactic* restriction on them, not a *semantic* one, which is all that
> > can be stated in DAML+OIL.
> 
> hmm, that's not my expectations.
> 
> e.g.
> 
> if lists are syntactic then
> 
> <daml:UniqueProperty rdf:about="http://example.org/#uniq"/>
> <rdf:Description rdf:about="#foo">
> <eg:uniq rdf:parseType="daml:collection">
>    <rdf:Description rdf:about="#member/>
> </eg:uniq>
> <eg:uniq rdf:parseType="daml:collection">
>    <rdf:Description rdf:about="#member/>
> </eg:uniq>
> </rdf:Description>
> 
> is contradictory and/or meaningless because the two lists are required to
> be semantically the same thing by the unicity constraint, but they are not
> (even though they are identical things - this identity is a semantic
> condition that could be defined with say graph isomorphism).
> 
> It seems to me that we have no mechanism in any semantic web technology to
> specify syntactic constraints on the graph; and so introducing one seems
> high cost.

The problem is that the RDF vision is extremely ill-suited to represent, or
even encode, syntax.  If you want to use RDF for OWL syntax, then the
constructs you want are just not available.  OWL itself does not make the
situation any better---OWL itself is ill-suited to represent, or even
encode, syntax.  So, yes, current semantic web technology is lacking in an
important area.  

Is this a problem?  I do not consider it to be, because I don't want to use
semantic web technology to describe the syntax of OWL (or RDF).  However,
it does become a problem when others attempt to do this.

> I would like to have a better understanding of the benefits of a syntactic
> mechanism that appears to me to merely duplicate semantic mechanisms that
> already exist.

No.  The mechanisms in RDF and OWL are very different from the mechanisms
one might want for representing syntax.  

Here are two examples:

1/ RDF M&S has a couple of places where it talks about syntactic
   restrictions on RDF documents (e.g., reified statments have exactly one
   subject, predicate, and object).  These restrictions cannot be captured
   in RDF, or RDFS, or OWL.  

   [I think that these restrictions may have been inadvertently dropped by
   the RDF Core WG.  I wonder how they will explain a statement with two
   subjects, or a collection with two first elements.]

2/ XML Schema is concerned with the syntax (and, to some extent, the
   semantics) of XML documents.  It thus has a very different flavour from
   RDF Schema, as is, of course, only appropriate.

The whole point of my messages in this thread is that the semantic
mechanisms from RDF and OWL are not appropriate for representing the syntax
of OWL.  That is fine by me, I don't really want to use them for this
purpose.  However, some people seem to think that the constructs of RDF and
OWL are fine for representing syntax.

If you want to represent lists as semantic objects in your domain of
discourse, then an OWL ontology for the DAML+OIL list constructs might be a
good way to go, or maybe not.  However, this would miss the point as far as
OWL (or RDF) syntax is concerned.

> Jeremy

peter

Received on Thursday, 2 May 2002 10:51:07 UTC