Re: DAML+OIL semantics

From: Jim Farrugia <jim@spatial.maine.edu>
Subject: Re: DAML+OIL semantics
Date: Tue, 20 Aug 2002 10:49:52 -0400 (EDT)

> Peter,
> 
> Some brief comments below.
> 
> Thanks for your detailed replies.
> 
> Jim
> 
> 
> [...]
> 
> > > JF:    OK, let's see....
> > > 
> > > 
> > >        1) what is the right way to get at the question whether the DAML+OIL 
> > >        axiomatic semantics "gives us the same thing" as the DAML+OIL 
> > >        model-theoretic semantics?  Does it make sense to try to ask such
> > >        a question, or is it a case of apples and oranges?
> > 
> > This is a valid question.  To be more precise, the kind of question that
> > you can ask is, given the axiomatization, A, and two DAML+OIL KBs, K and Q,
> > whether 
> > 	K entails Q in the model theory
> > iff
> > 	K + A entails Q in the model theory of the logic underlying the
> > 	axiomatization
> > 
> > One can also ask an equivalent question, namely whether 
> > 	K entails Q in the model theory
> > iff
> > 	Q can be derived from K + A in a sound and complete proof theory
> > 	for the logic underlying the axiomatization
> > 
> > Unfortunately, the answer to this question is no.
> 
> 
> The fact that the answer is no seems like not a good thing, though it 
> seems it may be a fact of life.  ??

Definitely *not* a good thing.

> To what can this fact be attributed - properties of the DAML+OIL 
> axiomatization? Properties of the logic underlying the axiomatization
> (which is what, by the way?)?
>
> Something else?

There are two reasons:

1/ The two semantics have differing treatments of the relationship between
DAML+OIL and RDFS.  The axiomatization tries for a much closer
relationship.

2/ There are a number of consequences of the axiomatization that are not
obvious from the language.

> How do DAML+OIL people respond to or deal with this fact, which seems
> to say that one can't be sure that the axiomatic semantics for DAML+OIL
> "give the same thing" as the model-theoretic semantics?

Well, most of us are ignoring the problems, and hoping that OWL will be 
different in some way.

> The reason I ask is that an implementer without much logical background
> might suppose that since two well-defined semantics are provided for
> DAML+OIL, they do in fact "give the same thing" as far as meanings and
> inferences are concerned.

Agreed, and the fact that there are two different co-normative semantics
with different properties is a serious problem.

> Thanks,
> 
> Jim

peter

Received on Tuesday, 20 August 2002 11:13:35 UTC