- From: Sandro Hawke <sandro@w3.org>
- Date: Wed, 22 Dec 2004 13:35:43 -0500
- To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
- Cc: www-rdf-logic@w3.org
> OK. However, the (best) part of FOL is entailment. If you haven't > captured that, then you havn't captured FOL, and it didn't seem as if that > is what you were going. I think I gave that mistaken impression by trying to keep my options open. My "uncle" example certainly involves entailment. > > I would also like the whole exercise to be, in the end, simple enough > > that the community interested in rule languages and RDF can be > > comfortably with the results. I keep hearing about projects using RDF > > encodings of Horn or FOL, and they turn out to have either unclear or > > inconsistent semantics. If we're going to move to a standard here, it > > would help to clear up whether it is possible to do this kind of > > encoding properly. I can't imagine, for instance, that the approach > > taken by SWRL 0.6 would ever be approved in a W3C Recommendation, for > > the reasons I outlined in the team comment [1]. > > Well, here we have a distinct disagreement. My view is that even if you > actually succeed, the result will not be worthwhile. > > Consider the case of OWL. Through much effort, OWL is a same-syntax > extension of RDFS. However, there are differences between this version of > OWL and the DL (i.e., more-or-less FOL) version of OWL, which is not great. > I expect that even a (mostly) successful effort of encoding FOL in RDF will > have similar divergences from "the one true FOL". It's possible, but those divergences might not matter in practice. I'd rather not throw away this chick before it's hatched. > Further, writing parsers for the RDF/XML syntax of OWL is a royal pain. > (Yes, I did say parser. An RDF/XML system is really only a tokenizer for > OWL.) Having to write such parsers eliminates any potential benefit > associated with having one Semantic Web "language". This "parsing" is the same as graph matching, which is the bread and butter of RDF applications, so I don't see it as any particular challenge. Maybe I'm missing some additional challenge -- I never read Sean's paper, and my OWL implementation just did little graph matches on the triples as needed. Trying to get the full "abstract" syntax out would be more challenging, I suppose, although probably not in a rule language. > > If Peter can convince me that it's not possible to write the desired > > semantic condition, I can help him convince others that a different > > path needs to be taken for rule languages. If I can convince Peter it > > *is* possible (and we figure out how onerous it is), then a standard > > can use it (or not, but because the difficulty has been actually > > weighed). > > > > Unfortunately, I can't just hand Peter the desired semantic > > conditions. I don't really know how to write them in the formalism of > > RDF Semantics [2]. My understanding is that Peter doesn't either and > > suspects it's not possible to write them. > > Well, I at least worry that it is not possible to do this extending in a > reasonable manner. (I certainly don't view using strings to encode > sentences as being in a reasonable manner.) Again, I don't want to prejudge the result. Mostly right now I'd like to find out of its theoretically possible to do; if it is, then we can try to see if there's a user-friendly way to do it. Encoding in strings might actually be very nice for users, and isn't a big problem for my style of formalizing the semantics, if we define a predicate connecting string literals to an rdf:List of the one-character string literals which form it. (It's tedious to map the strings to the triple form I've been using in examples, but it's mechanical -- it's basically a prolog DCG.) Sorry I'm not giving you a proper detailed answer to some of your questions yet; it's at the top of my non-interupt queue. -- sandro [1] http://lists.w3.org/Archives/Public/www-rdf-logic/2004Dec/0016
Received on Wednesday, 22 December 2004 18:32:51 UTC