Re: Axiomatic Semantics of OWL

[edited for length]

From: Richard Waldinger <waldinger@ai.sri.com>
Subject: Re: Axiomatic Semantics of OWL
Date: Mon, 18 Oct 2004 16:30:36 -0700

> Peter: 
> 
> It is certainly still necessary for us to develop a translator from OWL 
> syntax into the language of the axiomatic theory before the axioms can 
> be generally useful.    The axioms are for the Full version of OWL.

Hmm.  I'm not asking for a translator, just information on the translation,
without which your axiomatization cannot be checked for correctness.
As the translation is not an obvious one, you need to provide full
details on it.

If your axioms are for OWL Full, then your translation needs to work
(correctly) on all RDF graphs.

> I've interspersed some answers to your detailed questions below:
> 
> Peter F. Patel-Schneider wrote:
> 
> >From: Richard Waldinger <waldinge@AI.SRI.COM>
> >Subject: Axiomatic Semantics of OWL
> >Date: Sun, 17 Oct 2004 21:22:23 -0700
> >
> >>A validated set of first-order axioms for OWL, with discussion, is 
> >>available at
> >>http://www.kestrel.edu/home/projects/DAML/owl/axiomatic.html

[...]

> >I gave the axioms a quick look.  No inconsistencies jumped out at me, but I
> >had a very hard time figuring out what is going on.
> >
> >The biggest problem is that there is no definition of how the language of
> >the axiomatization relates to the OWL language (any variety), and I was
> >unable to make a guess at this relationship.  I was even unable to
> >determine which dialect of OWL should be the source of the translation.
> >Without this information the axiomatic semantics is not particularly
> >useful.  I have lots of other potential problems with the axiomatization,
> >but I can't determine whether these are true problems without knowing the
> >translation.
> >
> >Here is a very simple example illustrating the difficulty.  How do I
> >translate the following graphs into the language of the axiomatization?

[...]

> >Graph 2:
> >	ex:John ex:hasParent ex:Susan .
> >  
> >
>                   HasParent(John, Susan)
> 
> or , equivalently,
> 
>        Holds(hasParent, John, Susan)

Huh?  I don't see any axiom in your axiomatization that provides for the
equivalence of these two facts, even if the capitalization of hasParent
were the same in both.

[...]

peter

PS:  Here are a two more KBs to translate:

KB1:
	_:a owl:onProperty ex:p .

KB2:
	ex:a rdfs:subClassOf rdf:type .
	ex:John ex:a ex:Person .

Received on Tuesday, 19 October 2004 06:50:25 UTC