Re: ISSUE: DAML+OIL semantics is too weak

From: "Jos De_Roo" <jos.deroo.jd@belgium.agfa.com>
Subject: Re: ISSUE: DAML+OIL semantics is too weak
Date: Mon, 20 May 2002 23:16:12 +0200

[...]

> > How is modus tollens unsafe?
> 
> I think the trouble starts with your rule
>   ?x a ?y, ?x a [owl:complementOf :y] -> FALSE
> 
> We can't assert such a rule with a conclusion
> that is FALSE (as it is not a Horn clause).

It is a Horn clause, just not a definite clause, as you yourself have admitted.

| From: jos.deroo.jd@belgium.agfa.com
| To: pfps@research.bell-labs.com
| Cc: sandro@w3.org, phayes@ai.uwf.edu, www-rdf-logic@w3.org
| Date: Tue, 30 Apr 2002 21:47:27 +0200
| Message-Id: <OFCBB60AED.53261F74-ONC1256BAB.00686C6E@bayer-ag.com>
| 
| [...]
| 
| PS indeed a Horn clause is a disjunction with at most one positive literal
|    but we never assert clauses with no positive literal (which is a query)

I'm not responsible for the problems that you are getting yourself into by
only allowing definite clauses.  I don't have any problems with rules of
inference like modus tolens.  If you have problems with modus tolens
perhaps the problems are not with modus tolens but with your system.

> We would rather write a rule like
>   { :rule16c1 . ?x a ?y . ?x a ?z . ?y owl:complementOf ?z }
>   log:implies { ?x :inconsistentWith owl:complementOf } .
> 
> so have the inconsistency detected explicitly
> and that would not mean discharging assumptions.

I am unsure as to whether an approach like this would be viable.

peter

Received on Tuesday, 21 May 2002 14:18:28 UTC