Re: OWL Lite semantics

(This message follows those by Peter, Pat and Ian)

The formalists heard from, here's my "scruffy" response

EASE of implementation:

What Jeremy suggests is quite appealing to me because there is good 
evidence that it IS easier to implement -- in fact, the code foe a 
program with a considerable subset of Jeremy's OWL Lite proposal can 
be found in the 1987 AI textbook "AI Programming" by Charniak, 
Riesbeck, McDermott and Meehan *.

Current implementations:

In fact, a great many implemented AI systems use essentially what 
Jeremy proposes -- these are the systems that gre primarily from the 
frame systems (remember that discussion from very early in our 
group??)  - examples of a number of these tools can be found in the 
1999 "Ontology Management Workshop" that AAAI ran [1].

In these systems, the prevailing use has NOT been classification, and 
I'd argue that most of the "only if"s grow out of the need for 
classifiers (this is a simplification, I admit, but would take much 
more bandwdith to be specific).  However, in many cases people build 
the ontology separately from classification (thus not needing most of 
the class reasoning in OWL).

For example, in all the tools built by my current group [2], we 
assume that either we are importing an ontology from elsewhere, or 
that someone is extending one by adding classes and properties to 
existing places in an ontology.  In these cases, we are able to 
support many of the "if" entailments but not the only-ifs (c.f. my 
mexican restaurant example summarized by Deb in [3])

In short, under Jeremy's subset many current systems will already be 
able to handle OWL Lite, under our current definition they can't. 
(btw, everyone of the systems that I know of include HAS-VALUE, but 
only with the "if" semantics)


Pat and Ian claim Jeremy's proposal is somehow significantly 
different semantically than the current one.  This confuses me - what 
Jeremy proposes is still a proper subset of OWL DL, so I'm not sure 
why this one is so different from our current as to make trouble - 
can one of you explain?

* - sorry, this book predates the web and I can't find the code on the web.

Professor James Hendler
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)

Received on Monday, 9 December 2002 08:37:55 UTC