- From: <jos.deroo.jd@belgium.agfa.com>
- Date: Fri, 20 Jul 2001 01:54:15 +0100
- To: phayes@ai.uwf.edu
- Cc: w3c-rdfcore-wg@w3.org
>>[...] >>>>based on what we have testcased at >>>>http://www.agfa.com/w3c/euler/#Problem >>>>I completely agree with you and I really >>>>wonder how else we could model/infer >>>>http://www.agfa.com/w3c/euler/tpoint-result.n3 or >>>>http://www.agfa.com/w3c/euler/tpoint-all.n3 >>> >>>The answer is that in RDF, you *can't* infer this stuff. >>> >>>You use N3, which is way more expressive than RDF; it has negation >>>and universal quantifiers, just for a start. >> >>Fine, but per http://www.w3.org/2000/10/swap/log# >>and e.g. http://www.w3.org/2000/10/swap/model.n3# >>one *can* produce/consume regular N-Triples, no? > >Yes and no. They do produce triples, but how are those triples >interpreted? If they are interpreted using the meanings suggested by >the M&S (I refrain from saying 'semantics') then you can't validly >infer this stuff. Cwm infers it by imposing some extra >interpretations on the triples, ie in effect treating them as >triple-encodings of expressions in some other language, which as far >as I know has never been precisely defined anywhere, though it seems >to be something close to prolog with a nonstandard negation (? A >guess, as the only way to know for sure is to inspect the Cwm code >and reverse-engineer the intended semantics.). That is indeed true. We actually have to do a lot of effort to achieve the right interpretation. I think it can never be achieved in ONE component. We have almost no interpretation capability in e.g. proof engines but their results can be interpreted by other components which can cycle back new matter. -- Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/
Received on Thursday, 19 July 2001 19:54:51 UTC