- From: Axel Polleres <axel.polleres@uibk.ac.at>
- Date: Thu, 23 Jun 2005 00:35:59 +0200
- To: Jos de Bruijn <jos.debruijn@deri.org>
- CC: Jim Hendler <hendler@cs.umd.edu>, dreer@fh-furtwangen.de, public-sws-ig@w3.org
Jos de Bruijn wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Jim Hendler wrote: > > >>Jos- >>that's not quite right as I understand it - to use an example from the > > paper I cited - > >>Given an ontology containing only a single RDF triple: >><#pat> <#knows> <#jo>. >>the answer to a query asking if pat knows exactly one person would be > > "no" under > >>RDF's open world semantics, but "yes" under the closed world semantics > > of Datalog. > >>there's no negation explicit here. > > > If you would write down the query you mentioned, you would see that > the query includes negation. > > Fact is that a First-order theory: > > knows(pat,jo) > > entails only the fact: > > knows(pat,jo) > > > A Datalog program: > > knows(pat,jo) > > entails only the fact: > > knows(pat,jo) > > >>btw, this is not a small issue to me -- without some way to control > > default reasoning (i.e. limit the scope of axioms and facts that are > considered for a set of queries) there is just no way to link things > reliably on the Web (and without the linking, why bother with Web). > >>As far as I'm concerned, the minute one assumes there is a mechanism to > > close a graph (and several are floating around), then ontology and rules > go together just fine and lots of the layering is fixed. On the other > hand, if a rule can be stated in general that has a negation as failure > then we have a problem. > >>In the old days, for example my thesis work, we used "ThNOT" in > > langauges like MicroPlanner, NASL, etc. and that was some of the > earliest work on this stuff. The idea was you could use a rule like this > >>P :- ThNOT Q. >>to mean P is true if you cannot prove Q. Implicit in this, and many of > > the systems since, was the idea that there existed "A knowledge base" > which was all the facts that this rule could be used on. > >>On the Web, I would want this to read something like >>P :- (Thnot Q) given R. >>where "given R" somehow means with respect to some database, some > > graph, or some other nameable and definable entity. > >>Thus, for example, one could imagine running a query against some graph > > (and note in theory that this could be the identity query, meaning it is > equal to everything in the graph) and then stating that some set of > rules is applied with respect to the resulting graph. These rules having > NAF would be great and good and wonderful. If I applied your rules to a > different graph, it would be up to me to decide if I wanted to restrict > things in my graph to control the scope of the application of your rules > -- this sounds like a powerful and crucial mechanism, and with it a lot > could be done. Without it, rules on the web seem to me to either be just > doing logic programming with a consistent syntax (in which case why > invent all these languages, let's just use Prolog on the Web) or > pretending the whole Web can be treated like a closed world - and > that's, in my mind, where the dragons are... > > It indeed seems like a good idea to define scoped default negation. We > did not address this issue in WRL, but it would be a good topic to > address for the W3C rules working group (if it's going to be formed). > > > Best, Jos I could not agree more and think that this could be a very useful extension of our current framework. Problably most of you need not to be pointed to a related thread some time ago on this list, starting with http://lists.w3.org/Archives/Public/public-sws-ig/2004Jan/0040.html although it might be interesting to follow for those who did not... best, axel -- Dr. Axel Polleres Digital Enterprise Research Institute - DERI Innsbruck Institute of Computer Science, University of Innsbruck +43-512-507/6486 Axel.Polleres@deri.org http://homepage.uibk.ac.at/~c703262/
Received on Wednesday, 22 June 2005 22:36:14 UTC