W3C home > Mailing lists > Public > public-owl-dev@w3.org > July to September 2007

Re: declaredAs

From: Pat Hayes <phayes@ihmc.us>
Date: Fri, 10 Aug 2007 12:33:41 -0500
Message-Id: <p06230905c2e24e7525e1@[10.100.0.67]>
To: Alan Ruttenberg <alanruttenberg@gmail.com>
Cc: Bijan Parsia <bparsia@cs.man.ac.uk>, "Michael Schneider" <m_schnei@gmx.de>, public-owl-dev@w3.org, bmotik@cs.man.ac.uk

>Thanks Pat,
>
>Followup question in line
>
>On Aug 9, 2007, at 1:41 PM, Pat Hayes wrote:
>
>>>Does it break it, or does it make the OWL-1.1 entailments only be 
>>>a subset of OWL-Full entailments.
>>>If it breaks, could you say how please?
>>
>>Peter Patel-Schneider has made the point. RDF semantics is in line 
>>with CL semantics by interpreting names as individuals, and 
>>allowing individuals to act as classes and properties by mapping 
>>them to class and property extensions. This has the consequence 
>>that A=B means they are equal in every way, including as classes 
>>and as properties. 1.1-style punning maps each name to individuals, 
>>classes and properties separately, with no semantic association 
>>between these three denotation mappings. This allows the names A 
>>and B to have any pattern of identity, including being equal but 
>>not equal as classes, which is impossible in RDF or CL.
>
>In RDF this isn't possible anyways, is it, since there is no way to 
>express that two individuals are the same, is there?

True. But there is in OWL-level languages and above, so the question 
arises of whether equality is congruent with what the RDF semantics 
says is identity.

>CL is a point of interest, but I'm not sure how germane it is to a 
>discussion of OWL.

Well, I have a side axe to grind here, in that I want the W3C 
standards to be at least compatible with the CL ISO standard. Quite a 
lot of people in Government agencies are very keen to use an ISO 
standard when one is available, so it would I think be a mistake to 
break the link to CL casually, if it is easy to keep it by a 
relatively small change. But also, we have found in several recent 
projects that the CL semantic (and syntactic) 'style' has a lot of 
operational advantages. And more basically still, to me, the idea 
that things can be the same thing yet at the same time different 
classes, just seems semantically incoherent. It doesn't make 
intuitive sense. The fact is, one simply cannot have a logical 
equality in the punning semantics: but equality is a very basic 
notion, and is widely thought of as part of the fundamental apparatus 
of logic like the quantifiers or connectives. So the fact that it is 
incommensurate with the punning semantics strongly suggests that 
there is something basically wrong with the latter. OK, I concede 
that this is only a theoretician's intuition, but it is a very strong 
one, and I tend to rely on intuitions like this.

>
>Also, would this situation (A=B misses entailment 
>equivalentClasses(A,B) and equivalentProperties(A,B)) not fit my 
>characterization that the relation of OWL 1.1 to OWL Full is that 
>OWL 1.1 is missing some entailments that would be concluded using 
>OWL Full semantics?
>
>>One cannot build the 'punning' style semantics as a semantic 
>>extension of the CL-style model theory. One also cannot have 
>>genuine logical identity in the punning style semantics. (You can 
>>imitate it in OWL 1.1 by saying that it means (sameIndividual AND 
>>sameClass AND sameProperty), but this hack won't extend to richer 
>>logics and so will have to be redefined over and over again.)
>
>I think this answers the reverse question, namely can OWL 1.1 be 
>considered a semantic extension of OWL Full. I was asking if 
>OWL-Full can be considered a semantic extension of OWL 1.1.

Yes, it can. If one imposes the extra condition that sameIndividualAs 
always entails all the other identities, then the punning semantics 
becomes isomorphic to the RDFS/CL-style semantics. But this makes OWL 
1.1 into a kind of sub-logic rather than a real logic. I think its 
important to get the basic logic of the Web as intuitively 
semantically clear as possible, independently of considerations of 
efficiency for reasoners. If it turns out that nobody at Manchester 
or anywhere else knows how to make an efficient, complete 
theorem-prover for the language, then let them publish the 
deficiencies of their software clearly and tell people how to avoid 
them: but don't warp the basic communication language of the SWeb to 
fit the current state of the inference-engine art (which is in any 
case a constantly moving target.)

Well, now you know what I think, anyway :-)

Pat



>
>Thanks,
>Alan
>
>>
>>Pat
>>
>>>-Alan
>>>
>>>On Aug 9, 2007, at 12:23 PM, Pat Hayes wrote:
>>>
>>>>Not really. Im sure it was meant to have this intention, but the 
>>>>effect of moving to punning is two-fold: it breaks the OWL Full 
>>>>semantics, and it breaks the semantic connection between OWL and 
>>>>RDF. Neither of which are desirable, IMO, though both of them are 
>>>>in line with a certain perspective that has long been associated 
>>>>with Manchester
>>
>>
>>--
>>---------------------------------------------------------------------
>>IHMC		(850)434 8903 or (650)494 3973   home
>>40 South Alcaniz St.	(850)202 4416   office
>>Pensacola			(850)202 4440   fax
>>FL 32502			(850)291 0667    cell
>>phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes


-- 
---------------------------------------------------------------------
IHMC		(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32502			(850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
Received on Friday, 10 August 2007 17:33:55 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 10 December 2014 20:07:17 UTC