W3C home > Mailing lists > Public > www-webont-wg@w3.org > April 2002

Re: SEM: Face-to-Face version of approaches document

From: Jeff Heflin <heflin@cse.lehigh.edu>
Date: Fri, 05 Apr 2002 11:52:34 -0500
Message-ID: <3CADD652.240EEEBF@cse.lehigh.edu>
To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
CC: www-webont-wg@w3.org
Peter,

Thanks for a great document that clearly presents the choices we have
before us. Since I will be unable to attend the F2F, I'd like to provide
my initial votes on these issues.

1) I vote for B, a non-RDF XML dialect. My arguments can be found in
various threads on the mailing list. [1] [2]

2) I don't fully understand the implications of the issue, but I find
the particular example undesirable. Could you describe the basic issue a
little more formally?

3) I think this one oversimplifies things a little bit. Our requirements
document has some things that are not typically in a description logic
(e.g., equivalence of individuals, local unique name assumptions).
However, if we see this choice as sort of the fundamental part of the
language that may have some additional semantic extensions, then I
suppose I'd have to vote for D, roughly the power of DAML+OIL.

4) I am not in favor of formal sub-languages. If the Semantic Web
follows the "layer cake" design, there will already be plenty of layers;
if we starting adding sublayers to each layer, then things just get too
confusing (I've already heard people complain that they have to read too
many specs to understand this Semantic Web stuff: XML, RDF, RDF-Schema,
DAML+OIL, that's a lot of reading!). However, I would be in favor in
using the layers informally to present the language. This allows people
to learn the basics quickly, and to gain proficiency in the language at
their own pace.

5) I think we need to distinguish between RDF model theory and RDFS
model theory here. Also when you say in B that the model theory might be
"close to RDF," I assume you mean it is an extension that is almost
compatible. Assuming we go with a non-RDF syntax for ontologies, then I
would vote for a model theory that is almost or fully compatible with
RDF for instances, but is different from RDF Schema for ontologies.

6) I'm not sure about this one. I am leaning in favor of it, because it
appears to me to be simple entailment. If we translate this into FOL
with unary classes for predicates, then this seems to say that:

Student(john)
Employee(john)

entails

Student(john) AND Employee(john)

I can't see how we could have a semantics that doesn't say that. But
since Pat Hayes is arguing against this, and he is more of a logic
expert than I could ever hope (or want ;-) ) to be, I must be missing
something here.

Jeff

[1] http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0277.html
[2] http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0337.html

"Peter F. Patel-Schneider" wrote:
> 
> I updated
> 
> http://www-db.research.bell-labs.com/user/pfps/webont/approaches.text
> 
> in response to the comments received during the teleconference.
> 
> This is the face-to-face version.  However, comments before the
> face-to-face are still welcome, and will be addressed at the face-to-face.
> 
> peter
Received on Friday, 5 April 2002 11:52:37 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:49 GMT