Re: Issue: Add hasValue to OWL Lite

On October 31, Jim Hendler writes:
> 
> At 3:45 PM +0000 10/31/02, Ian Horrocks wrote:
> >On October 29, Dan Connolly writes:
> >>
> >>  On Tue, 2002-10-29 at 18:34, Ian Horrocks wrote:
> >>  >
> >>  > The problem with adding hasValue to OWL Lite is that it wouldn't be
> >>  > Lite any more. The lack of hasValue in Lite is, from an implementation
> >>  > point of view, the main thing that differentiates it from fast -
> >>  > hasValue is very tough to deal with, and is responsible for pushing
> >>  > the worst case complexity of reasoning in fast OWL from ExpTime to
> >>  > NExpTime.
> >>
> >>  Could you unpack that a bit?
> >>
> >>  Could you give an example, maybe?
> >
> >I'm not sure. This isn't anything to do with reasoning techniques or
> >specific examples, it is a fundamental property of the logic that
> >basic inference problems (satisfiability, subsumption, entailment) are
> >much harder when we add extensionally defined classes (which is what
> >hasValue amounts to).
> >
> >If you want an intuition, it comes down to the loss of the tree(ish)
> >model property. Without this property, it is very hard to devise
> >decision procedures that work in a goal-directed way and that know
> >when they are done.
> >
> >Ian
> 
> Ian - the question arose at the Telecon as to whether this was true 
> for both the IF and the ONLY IF (i.e. hasValue -> X vs X -> hasValue) 
> -- that is, does saying "All Mexican restaurants serve Mexican food" 
> cause the problem if you're not expected to be able to say "all 
> places that serve Mexican food are Mexican restaurants"??

I'm not sure I understand the question. In particular I'm not sure
what is meant by "not expected to be able to say..." - do you meant
that the language doesn't allow you to say it or just that you don't
expect that users will say it? As you can imagine, I wouldn't be very
happy with any argument based on the latter interpretation.

As far as the iff is concerned, it isn't easy to see how we can avoid
this possibility: we can assert that some class C is sameClassAs a
hasValue restriction, and then we can use C on either the left or
right hand side of other axioms. 

In any case, nominals can cause lots of nasty effects even without
using them on the lhs of axioms. E.g., by using an inverse-functional
property in a hasValue restriction the cardinality of a class (even of
Thing) can be restricted to 1. As we have seen before, finite domains
can lead to unexpected (and hard to find) inferences.

Ian


>   -JH
> 
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> 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)
> http://www.cs.umd.edu/users/hendler

Received on Thursday, 7 November 2002 06:32:42 UTC