Re: Examples of OWL used for datatype inferencing?

Graham Klyne wrote:
we have been using full first-order inference to reason about theories 
that combine OWL with stronger subtheories such as the natural numbers.  
we have done this earlier for daml+oil and are now working on a 
first-order axiomatization of OWL.  We use SNARK as the inference engine.

By going to first-order reasoning you lose decideability but gain 
expressive power.  OWL-full is not decideable anyway, I believe.

Richard

>
> [I originally asked this on RDF-IG, but realize this is probably the 
> better forum. #g]
>
> Does anyone have any examples of using OWL to perform 
> RDF-datatype-related inferencing?
>
> I'm thinking of datatypes, such as numbers, for which additional 
> properties are used to define additional relations, such as addition 
> over numbers.
>
> For example, given:
>
>    :vehicle :seatedCapacity "30"^^xsd:integer .
>    :vehicle :standingCapacity "10"^^xsd:integer .
>
> and knowledge that the total capacity is seated capacity + standing 
> capacity, that one might infer:
>
>    :vehicle :totalCapacity "40"^^xsd:integer .
>
> This might be expressed thus using CWM-style rules:
>
>    { ?v :seatedCapacity ?c1 .
>      ?v :standingCapacity ?c2 .
>      (?c1 ?c2) math:sum ?c3 . }
> =>
>    { ?v :totalCapacity ?c3 . }
>
> It seems to me that to express such relations one must have a form of 
> universal quantification.  But I'm not sure if anything in OWL 
> performs such a purpose, so I struggle to see how one might express an 
> idea like that above.
>
> Behind this question, I'm trying to see if there's a way to abstract 
> the rules of datatype properties away from particular application 
> domain.  (i.e. using just RDF statements, and not rules, to express 
> ideas like the example above, appealing only to 
> application-independent rules defined on datatyped values.)  Currently 
> I'm not seeing any way to do this, but before I give up I wanted to 
> see how OWL (as the major thrust for Sweb inference) deals with such 
> issues.
>
> #g
>
>
> ------------
> Graham Klyne
> For email:
> http://www.ninebynine.org/#Contact
>

Received on Friday, 31 October 2003 13:40:29 UTC