- From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
- Date: Tue, 19 Mar 2002 10:02:07 -0500
- To: phayes@ai.uwf.edu
- Cc: www-webont-wg@w3.org
From: Pat Hayes <phayes@ai.uwf.edu> Subject: Re: Moving forward Date: Sun, 17 Mar 2002 12:38:08 -0800 [...] > >A fourth way to go would be to > >1/ use a different semantic relationship for inference > > I do not understand what this means, and request clarification. > Entailment is pretty much established by a semantics, seems to me: A > entails B if every model of A is also a model of B. That doesn't > leave much flexibility for different semantic relationships for > inference. The sort of thing I am alluding to here is similar to what I see in some of the recent messages. For example, to determine if i rdf:type :_1 :_1 ..... follows from a KB, don't use entailment directly. Instead, augment the KB with the appropriate triples related to :_1, and then ask whether the augmented KB entails i rdf:type :_1 Admittedly, this kind of inference is based on entailment, but it is not exactly entailment. Another possibility is to have i rdf:type :_1 :_1 ... follow from a KB if in all models of KB i is an element of the class extension of :_1, where class extension is defined in a manner similar to the way it would be in a description logic. This kind of inference is *not* based on entailment, although it is, of course, closely related to entailment. peter
Received on Tuesday, 19 March 2002 10:03:41 UTC