- From: Michael Sintek <sintek@dfki.uni-kl.de>
- Date: Tue, 21 Nov 2006 17:07:10 +0100
- To: Dave Reynolds <der@hplb.hpl.hp.com>
- CC: Sandro Hawke <sandro@w3.org>, public-rif-wg@w3.org
Dave Reynolds wrote: > Michael Sintek wrote: >> >> Dave Reynolds wrote: >>> Sandro Hawke wrote: >>>> Thinking more about the abstract syntax question, I find myself >>>> wondering how an abstract syntax is really different from an ontology. >>>> Is there any practical difference between an ontology of Horn rules and >>>> an abstract syntax for Horn rules? >>> One possible difference is that in an abstract syntax you can say >>> whether ordering is significant or not. Sure, as you argue, one benefit >>> is that you can say "ordering is not significant here" but you can also >>> (at least in ASN.1) say "ordering *is* significant here". Or perhaps >>> that is just a feature of ASN.1 rather than of abstract syntax notations >>> in general. >>> >>> However, I think an ontology is closer to want we actually want for RIF >>> and the only places we need to preserve order (e.g. in function/relation >>> arguments) can be handled in an ontology. >>> >>>> If this is true, then OWL is a probably a good way to write down our >>>> abstract syntax for RIF. >>> That's also, sort of, what Michael Sintek was pointing out. >> >> With the big difference that I did not use OWL, since the open-world >> semantics of OWL can result in unexpected results (like missing parts >> of rules are automatically "inferred"). > > Understood, that's why I said "sort of". > >> What we need is a more or >> less closed-world, database schema like ontology language, so exactly >> what RDFS was originally meant to be but is not any more. > > Actually I think that is an issue of processing rather than of modelling. > > We can use OWL to model what we say will be true of a rule, e.g. that it > should have at least 1 consequent. > > It's true that if one then applied a generic OWL inference engine to it > one would infer the presence of a missing consequent rather than detect > an omission. However, it is perfectly possible to also take the same > modelling but apply closed world checking to instance data (we have > tools that do this) and so achieve the desired schema-like behaviour. IMHO, using OWL for modeling but interpreting it in a schema-like way is not allowed -- this would mean we completely ignore the official semantics. So if we send our RIF OWL ontology to other people who use the official semantics, they will understand it in a different way, and then we cannot use it as an interchange language any more. But, as Sandro already pointed out: we can *try* to use OWL with the correct semantics, find the cases where some unwanted things happen (if at all), and try to remedy them with additional OWL axioms. Michael > > In that way one could use asn06 is a convenient syntax, translate to OWL > to give a formal interchange, use that in closed world checkers for > validation and get an instance serialization syntax (RDF as you say) for > free. > > Dave > > > Dave > > -- Michael Sintek -- DFKI GmbH, Kaiserslautern http://www.michael-sintek.de -- sintek@dfki.uni-kl.de phone: +49 631 205-3460 -- fax: +49 631 205-4910
Received on Tuesday, 21 November 2006 16:07:42 UTC