Re: Cardinality Restrictions and Punning

Peter F. Patel-Schneider wrote:
> The examples from Jeremy and Michael show where punning is weaker than
> identity.
> 
> Many such examples can be had, including
> 
> 	SameIndividual(owl:Thing owl:Nothing)
> 
> which is inconsistent in OWL 1.0 Full but consistent in OWL 1.1 DL and
> in OWL 1.0 DL.
> 
> What is new in the examples is that OWL 1.0 DL did not allow punning
> between data and object properties.  Aside from this expansion of
> punning, I don't see anything new here.
> 

Yes your example in OWL 1.1 DL does seem to me to show how confusing 
punning is. The class extension of owl:Thing and owl:Nothing are 
required to be different, so the fact that the ontology consisting of 
your single axiom is consistent does seem unhelpfully counter-intuitive.

I will ponder whether there is any significant difference with the 
property punning over the general point that punning gives OWL DL a 
weaker semantics than OWL Full.


My initial draft response was perhaps a little too annoyed:

  Your example from 'OWL 1.0 DL' strikes me at least as disingenuous.

  Yes, the spec does specify such a beast, but it does not specify how to
  write it down. If you wish to write down an OWL 1.0 DL ontology, it has
  to be done in RDF/XML, and to avoid punning as specified in section 4
  of S&AS; i.e. SameIndividual(owl:Thing owl:Nothing) has never been
  consistent in any ontology language used on the Web.

Jeremy

PS I note I made a small mistake, my example uses owl:DataRestriction, 
which is of course from OWL 1.1 rather than OWL 1.0. It is only by 
replacing it with an owl:Restriction that I get an inconsistent OWL 1.0 
Full ontology. It may be possible to tweak the OWL 1.1 Full semantics to 
somehow exploit this slight difference, and get the desired 
compatibilities. I don't hold out great hopes :(

Received on Friday, 21 December 2007 16:50:22 UTC