- From: Wolfram Conen <conen@wonkituck.wi-inf.uni-essen.de>
- Date: Thu, 29 Nov 2001 17:40:49 +0100 (MET)
- To: jbroeks@cs.vu.nl (Jeen Broekstra)
- Cc: www-rdf-interest@w3.org
Interesting issue. > I am also unsure whether this idea of flexible addition of > expressivity is feasible at all. The problem to me seems to > be that you need a starting point that is itself at least as > expressive as the language that you are trying to "learn", > which kind of defeats the purpose of the undertaking. Hm, what exactly is the purpose of the undertaking? Let me point out a simple possibily: why don't we agree upon one (or a set of) "host formalism(s)" that gives us the expressivity required to develop RDF vocabularies and RDF-based applications easily and interchangably - perhaps taken out of the datalog (or Ur-datalog) family. With a minimal hardwired transformation (for example: from triple to a 3-ary predicate), semantics of properties could then be defined "on the fly" (as needed by the application-domain community developing it) by giving a definition of the property embedded into RDF in a suitable codification of axioms -- this can be done with embedded RuleML or RDF-encoding of rules a la McDermott, for example (or a XML-serialization of S-KIF or whatever -- even alternatives in one document for different host formalism are possible). Then, we could start to develop semantically -augmented RDF schemata based on sufficiently expressive (though, admittedly potentially expensive) and well-known formalisms, we may even run into undecidability problems once in a while (depending on the chosen formalisms), but we would have something that is easy to use for a number of developers, and that can drive the evolution of RDF vocabularies and applications. (We could later still decide to reduce the possibilities again if that should be necessary). We can then use a mapping of DAML or RDFS to suitable formalisms and schemata that are augmented accordingly with axioms expressed in this formalism to restrict expressivity to some commonly agreed upon "language" -- but we can also easily add/change/augment semantics to application-specific needs, experiment with other subset-languages of the chosen host formalism on the fly, may even come to something as a competition of vocabularies etc. We can, of course, continue to reinvent the wheel of logic for RDF*/RDFS/RDF++ anew and keep the whole SemWeb-(r)evolution stalked a bit longer ;-) Ciao, Wolfram (*: Brief remark: If we do so, I am not sure if assuming that RDF is not a convention to speak in subject/ predicate/object patterns (with the possibility to interpret reification as a means to build nested structures etc.) but a logic language asserting instances of binary relations, is really helpful - but that's another topic). PS: of course, embedding axioms in RDF and "interpreting" RDF/RDFS inside of a host formalism, has been suggested earlier (Staab et al, Conen/Klapsing, I can give precise references if needed, simply send me an email), but I've never really seen a discussion of it. If I missed it, please, send me a pointer. Thanks.
Received on Thursday, 29 November 2001 11:40:51 UTC