RE: RDF and OWL rules

Yes Geoff, that very much does seem like a reasonable approach.
Modulo renaming, all exi-vars in the body do become uni-vars
within the scope of the rule body => head, so that could help
as well I think. Wrt to comprehension, I'm still thinking that
there is an alternative for Skolem functions, but I haven't
found it yet...

-- ,
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/


                                                                                                                   
                    "Geoff                                                                                         
                    Chappell"            To:     Jos De_Roo/AMDUS/MOR/Agfa-NV/BE/BAYER@AGFA                        
                    <geoff@sover.n       cc:     <www-archive@w3.org>                                              
                    et>                  Subject:     RE: RDF and OWL rules                                        
                                                                                                                   
                    2003-04-05                                                                                     
                    08:20 PM                                                                                       
                                                                                                                   
                                                                                                                   




Hi Jos,

> -----Original Message-----
> From: Jos De_Roo [mailto:jos.deroo@agfa.com]
> Sent: Friday, April 04, 2003 6:49 PM
> To: geoff@sover.net
> Cc: www-archive@w3.org
> Subject: RE: RDF and OWL rules
>
>
[...]
> > I wonder if an initial target of supporting lite-like cardinalities
(0,
> > 1, >1) might be a good idea for this reason.
>
> indeed, we also like "first doing the simple things"

On this note, I've been looking at breaking down the rules into the
following groups:

- rules arising from "if" semantics of vocabulary items (we should be
able to capture all of these. I think these are the ones that can be
expressed with all universally quantified variables in the head, and all
other body variables existential - a natural fit for LP languages)

- rules arising from "iff" semantics of vocabulary items (we can do some
of these. I think these amount to the rules with all universally
quantified variables in the head, but also some of the body variables
are universally quantified. I imagine we can only capture those that
allow us to evaluate a forall in the body without making a closed world
assumption)

- rules arising from comprehension principles (we can do some of these -
I think these are the ones that have some existentially quantified
variables in the head requiring a skolem function of some sort; much
danger of infinite generation here)

I've been categorizing the different rules along those lines (I'll send
you those when I'm done). I'm thinking that we ought to be able to close
off the first group pretty easily - we just make sure we have a rule
that expresses the if conditions of each vocabulary item (we probably
have all of those rules already)

For the second and third groups, I want to try to further characterize
the ones that we can capture and the ones that we can't. Hopefully based
upon that characterization, we can't be sure that we've gotten all of
the ones we can.

Does that seem like a reasonable approach to you?

Regards,

Geoff

[1] http://www.w3.org/TR/2003/WD-owl-semantics-20030331/rdfs.html

Received on Saturday, 5 April 2003 13:42:33 UTC