Re: OWL reasoning in rules

On 25 May 2007, at 13:48, Adrian Walker wrote:

> Hi All --
>
> As usual, Matt asks the good questions.
>
> Actually, you can to a large extent have your expressiveness cake  
> and eat it.
>
> The theory foundation for this is in the paper [1], and [2] is an  
> overview.
[snip]

I didn't know you had a full and correct reduction of OWL (full? dl?  
lite?) to IBL! I would have thought that this would have standardly  
involved an exponential blow up in the number of rules and thus be  
totally impractical (Boris's reduction can sometimes get slammed by  
that, and its targeting a logic of comparable worst case complexity  
to owl).

Are there any details of your reduction? I thought IBL was roughly  
Datalog, so pretty impractical as a reduction target for a decision  
procedure for OWL DL, but perhaps you've done something exceedingly  
clever? Are you planning to publish the details anywhere? It's  
certainly of great interest.

How do you do compared to Racer, Pellet, FaCT++ and esp. KAON2? Are  
there any benchmarks? Does your reduction do the sane thing with  
regard to DLP and hornSHIQ kbs?

Cheers,
Bijan.

Received on Friday, 25 May 2007 18:15:41 UTC