Re: exchanging OWL through RIF

---- Original message ----
>Date: Fri, 03 Mar 2006 14:44:32 -0500
>From: Michael Kifer <kifer@cs.sunysb.edu>  
>Subject: Re: exchanging OWL through RIF   
>To: bparsia@isr.umd.edu
>Cc: RIF WG <public-rif-wg@w3.org>
[snip]
>
>My statement should be understood in the context of Ed's use case.
>He conjectured that there might be an FOL rule engine that *doesn't
>understand OWL* and therefore a translation is needed. 

That's what Hoolet does. It translates OWL DL and SWRL into FOL and evalutes 
on the FOL theorem prover, Vampire. (I believe it generates TPTP format, so 
could use fairly arbitrary TPs). mSpass does something similar.

> SWRL
>is not such an FOL reasoner, 

Er...you mean rule language?

> because it builds directly on OWL.
>OWL doesn't need to be encoded in order to be used by SWRL.

This is true.

> It is the need
>for this encoding that I claimed to be questionable.

I largely agree. Especially as it's not that hard to do the direct translation, and 
for any kind of practicality, you'll need to do more elaborate translation anyway.

>Now, the charter talks about FOL rule languages and there will be a need to
>devise a RIF encoding for something like SWRL. But I doubt that it will
>look like a translation into FOL (one that extends the usual translation
>from DL to FOL).

Prolly not.

Bottom line is that there *are* (a few, experimental) systems that have 
proceeded via a translation of OWL DL to FOL, with surprisingly good results. 
(KAON2 also works this way, since it hits disjunctive datalog via the FOL 
translation...however, that's far far far from a transliterate and go approach.)

Cheers,
Bijan.

Received on Friday, 3 March 2006 20:04:05 UTC