# Re: OWL-DL as graph

From: Matt Williams <matthew.williams@cancer.org.uk>
Date: Tue, 20 Sep 2005 09:52:25 +0100
To: Jerome Euzenat <Jerome.Euzenat@inrialpes.fr>, W3RDFLogicList <www-rdf-logic@w3.org>
Message-Id: <1127206345.4847.18.camel@localhost.localdomain>

Dear Jerome,

Thanks for the reply, and my apologies for not making things clearer.

> I am not sure I understand the syntax here, can you paraphrase it?
> (how are the Cs quantified? are they constants? I assume the
> existential quantifer is on the role R).

1: "exists R(C exists R'.C')"

Where the existential quantifier is on the R (and R'), and C (and C') is
some Class in the T-Box.

>
> The two tracks would be to look for Feature-structure (feature
> algebras...) and Conceptual graphs (a.k.a, conceptual structures).
>
> However, both of them do not go easily with negation.

The issue here is that CGs are not equivalent to DLs (for a variety of
reasons).

I have started by defining a metric for the length of the expression,
and for the negated expression:

1: C ex. R(C' and ex. R'(C'' and ex. R''.C'''))
2: C not ex. R(C' and ex. R'(C'' and ex. R''.C'''))
3: C ex. R(C' and not ex. R'(C'' and ex. R''.C'''))
4: C ex. R(C' and ex. R'(C'' and not ex. R''.C'''))

This morning, I've shown that you could define the length of 1 as 2n-1
(where n=number of classes), to capture the number of classes and
properties. We can then define the length of 2-4 as being the length up
until the point of negation, normalised by the length of 1.

The problem is, this is fairly ad hoc (although it captures some useful
intuitions).

I have the feeling that I am reinventing the wheel here...but haven't
seen anything else that discusses this topic.

Matt
Received on Tuesday, 20 September 2005 08:52:41 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:50 GMT