RE: Classifying with Inequalities

Bob,
 
Jeremy's reply has clearly pointed out that there are efforts currently
underway.
 
Gary
 
-----Original Message-----
From: Bob MacGregor [mailto:macgregor@ISI.EDU] 
Sent: 10 May 2004 10:21
To: www-rdf-logic@w3.org; Gary Ng
Subject: RE: Classifying with Inequalities
 
If OWL classifiers are moving ahead of OWL, and adding in capabilities
based on practical needs, this is a good thing.  However, it does seem
to indicate that OWL is already behind the times, and validates the
perception that OWL was designed without paying much attention
to the demands of real-world applications.  It would
be very unfortunate if users have to resort to using language constructs
in Cerebra and/or Racer that do not conform to the OWL spec, since that
encourages a proliferation of non-regulated modeling constructs, which
is exactly the kind of thing that a language standard is supposed to
prevent.

Can anyone suggest if there is a possibility that OWL or SWRL could
standardize on a set of inequality properties within a reasonable time
frame,
so that we don't have to each invent our own, and then do alignments
down the road?
Are there semantic issues that need to be resolved, or are the semantics
sufficiently
obvious that we could just informally agree on them, and then put them
on a committee's
TO DO list?

Cheers, Bob


At 09:36 AM 5/10/2004, Gary Ng wrote:


Hi Bob,

Unfortunately classifying by datatype restrictions whereby each datatype
may be a user-defined range is currently not in the OWL specification,
but it is not incompatible. 

I believe there are two reasoners which has the capability of doing the
classification you have described: Cerebra and Racer. Cerebra uses an
extension of OWL to specify user defined ranges. 

I will leave the list to reply to the question of why OWL did not
include such features. But IMHO it was more logistics than
technicalities. It has something to do with the stability of RDF
datatypes and integration issues with XMLSchema datatype at the time, to
fully integrate such features would have been another project in itself.


Cheers

Gary

Gary Ng, Ph.D.           <gary.ng@networkinference.com>
Network Inference Inc.
5900, Laplace Court, Suite 250
Carlsbad, San Diego, CA 92008
Tel: +1 (760) 476 0650


-----Original Message-----
From: www-rdf-logic-request@w3.org [mailto:www-rdf-logic-request@w3.org]
On Behalf Of Bob MacGregor
Sent: 10 May 2004 09:11
To: www-rdf-logic@w3.org
Subject: Classifying with Inequalities


We do a lot of reasoning with inequalities, e.g., restricting
events within certain dates, or logging observations
within rectangular regions. It would be extremely
useful to be able to define classes that include such
restrictions and to arrange them in a classification hierarchy.
My impression is that there
is no inequality operator for any of the OWL variants.  My
question is, are inequalities compatible with OWL, or is this
yet another area (e.g., like metadata and property composition) 
where the existing OWL infrastructure falls far short of user needs?

In either case, it would be very nice if there were inequality
properties
for greater than, less than, greater-than-or-equal, less-than-or-equal
blessed by a W3C standard. Is there any possibility of this
happening in the near future?

Note:  Loom added the ability to classify scalar
intervals, and restrictions that reference them, very early on.  
Thus, the notion of the kind of classification I need has been
around for a very long time.

Cheers, Bob
=====================================
Robert MacGregor
Senior Project Leader
macgregor@isi.edu 
Phone: 310/448-8423, Fax:  310/822-6592
Mobile: 310/251-8488

USC Information Sciences Institute 
4676 Admiralty Way, Marina del Rey, CA 90292 
=====================================