Re: [UCR] RIF needs different reasoning methods

On Mar 10, 2006, at 3:58 AM, Francois Bry wrote:

> Bijan Parsia wrote:
>> What I don't see is the reason for specifying particular proof
>> procedures *instead of* expressive subsets. Responsiveness
>> requirements are application dependent, not document dependent (as far
>> as I can tell). That is, you want to *in the context of a particular
>> application* specify which reasoner, given certain parameters, to use.
>> If two reasoners perform acceptibly and give the same answers...what
>> *more* do you need?
> Let consider an analogy. Assume oner wants to insert a data item I in a
> collection C of data items and get the resulting collection sorted.
> Depending on whethner C is sorted or not, this can be done in different
> manners with considerably different complexities.
>
> Similar cases will arise with RIF applications. Drawing consequences
> from facts and deduction rules, checking if normative rules are
> satisified or not, or drawing consequences from general rules (or
> formulas) are three different tasks. The semantics of these tasks might
> well be expressable in the same manner. But performing these three 
> tasks
> at acceptable, state-of-the-art efficiency requires different methods.

Sure. But that's not an exchange issue. That's an application issue. Or 
service description issue.

> My suggestion is to make annotations (to RIF rulersets) possible using
> which one could specify such different tasks and thus giving hints at
> how to perform them efficiently.

Well, gee, I suggested that too. My question is whether the annotations 
are hints or not.

> I do not see any darwbacks in making
> such annotations possible. I see considerable drawnbacks in not making
> such annotations possible.

Well, the question is what hints are useful. I.e., what information 
needs to be captured and conveyed.

Cheers,
Bijan.

Received on Friday, 10 March 2006 17:02:48 UTC