- From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
- Date: Fri, 13 Oct 2000 23:22:13 -0400
- To: connolly@w3.org
- Cc: www-rdf-logic@w3.org
From: Dan Connolly <connolly@w3.org> Subject: DAML/RDF: a semantics and 2 more syntaxes Date: Fri, 13 Oct 2000 16:17:13 -0500 > In response to a recent request for a semantics for DAML: > > [...] > > I'll share my my understanding of RDF semantics, and > hence DAML semantics. Please understand that this > is my own personal view of DAML/RDF; Jim H. has > made it clear that DAML doesn't (yet) specify > any one semantics, and lots of folks (including > my co-authors!) have disagreed, in large or > small part, with the position I take; > to whit: > > [[[ > In [KIF] terminology, an RDF document is an atomic, > simple, ground, knowledge base, restricted to > 2-place predicates. > ]]] > > -- section 6. A Logic built on RDF-NF > of An Agent Markup Languagee version 0.5 draft > http://www.w3.org/2000/07/DAML-0-5#Building > > To elaborate slightly: an RDF document is a serialization > of a graph. To understand it, you parse the serialized > form into a graph, and then consider each arc > from s to o labelled p to be a KIF sentence (p s o); > pretty much the same as a prolog fact p(s, o). > p, s, o are URIs. > > In particular, I regard daml:equivalentTo as > having the same semantics as (= X Y) in KIF. > i.e. > > (defrelation http://www.daml.org/2000/10/daml-ont#equivalentTo (?x ?y) > := (= ?x ?y) ) OK, this is a potential specification of the meaning of equivalentTo via translation into some second-order logic. Under this specification, it appears that equivalentTo would allow making equivalences between unrelated logical statements. However, the translation from daml into kif provided in the attachments to your message doesn't carry this at all. In fact, I don't see ANY semantics being provided by this translation (and I'm not sure that is was supposed to, by the way). However, if it was intended that the translation into KIF syntax was not intended to provide some help for semantics, then I don't know what it was intended to be. > The recent suggestion[026] that defined classes can be > expressed without quoting/reification seems to > suggest that DAML consumers be allowed to > rely on information from the XML serialization about > which properties were stated where. If you > want to use RDF APIs and tools, that won't work. > This information disappears when an RDF document is > parsed into a graph. I don't see this at all. Again, let me reiterate my plea. What I want is some clear idea of what the meaning of the DAML-ONT constructs are. I don't care too much about how this meaning is conveyed, except that it be clear. (Of course, I do happen to think that some methods for conveying meaning are better than others.) To go along with the meaning, I would also like to see what restrictions are placed on the constructs (see the ongoing discussion of the difference between these two with respect to equivalentTo). Peter Patel-Schneider
Received on Friday, 13 October 2000 23:24:07 UTC