Re: ISSUE-52 (Explanations): Specification of OWL equivalences and rewriting rules for explaining inferences

On Nov 2, 2007, at 6:02 PM, OWL Working Group Issue Tracker wrote:

> ISSUE-52 (Explanations): Specification of OWL equivalences and  
> rewriting rules for explaining inferences
>
> http://www.w3.org/2007/OWL/tracker/issues/
>
> Raised by: Vipul Kashyap
> On product:
>
> I was wondering if the OWL 1.1 effort should also look at ways and  
> means of standardizing inference explanations, especially to make  
> them user understandable.

IMO, no.

> The current version of P4 has functionality that identifies the  
> relevant axioms involved in making an inference, but stops short of  
> explaining how the entailments/consequences of these axioms can be  
> chained together to create an explanation.

This is an active research topic. If you are coming to the face to  
face, I can show you some of that research.

You might also look at Swoop which has more advanced presentation  
features. The implementation in P4 is preliminary (e.g., not  
including any presentational help). You might be surprised how far  
you can get without lemma generation.

> Towards this end, I was wondering whether well know OWL/DL  
> equivalences and rewritings should be part of the OWL 1.1 Spec.  
> Some examples that come to mind are:
>
> A subClass B ==> (p some A) subClass (p some B)
> (p some (A and B)) subclass (p some A) and (p some B)
>
> I am sure there are many others ...
> A standardized approach for explanations of inferences could be a  
> very useful feature from the user point of view.

Actually, this is pretty clearly false. What would be very useful is  
very useful. I'm willing to bet a lot that the above rewritings are  
not. We already know (through actual user study, albeit indirectly)  
is that presentation is helpful. We have preliminary evidence that  
certain kinds of rewrite (e.g., swoop's "strike out" mode) can help a  
lot, and we are exploring others.

However, this is not appropriate for this group. Nor would it  
significantly speed access to users.

Cheers,
Bijan.

Received on Friday, 2 November 2007 21:06:19 UTC