relationship between RDF and DAML+OIL (from Pat Hayes)

Pat sent this message just to Jim and me.  I think that it is an excellent
discussion of some of the issues involved in the (tangled) relationship
between RDF and DAML+OIL.

(Either Pat forgot to cc the list, in which case I'm doing him a favour
here, or he thought that no one else would care, which I don't think is the
case.)

peter

Forwarded message 1

  • From: Pat Hayes <phayes@ai.uwf.edu>
  • Date: Tue, 11 Dec 2001 22:57:54 -0600
  • Subject: Re: proposal for working on the ontology language
  • To: Jim Hendler <hendler@cs.umd.edu>
  • Cc: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
  • Message-Id: <p05101005b83c898856c3@[65.212.118.193]>
>At 11:27 AM -0500 12/11/01, Peter F. Patel-Schneider wrote:
>>
>>
>>2/ Syntax
>>
>>    Changes to the model theory for DAML+OIL may make a different syntax for
>>    DAML+OIL more attractive, or, alternatively, make the current syntax
>>    less attractive.  I think that it would be better to divide DAML+OIL
>>    into two parts---the RDF part and the non-RDF part.  The RDF part would,
>>    I think, best retain the RDF syntax, but the non-RDF part might be
>>    better put in a different syntax.
>>
>
>Peter - this is an interesting idea, perhaps you could give a short 
>precis to the group as to what you have in mind.

Let me try also.  Peter, if I get anything wrong, correct me.

RDF consists of sets of triples. Sets of triples are not an adequate 
syntax for all of DAML+OIL. However, the syntactic structures needed 
to support DAML+OIL can be *encoded* in a triples format, and we have 
done this. However, some of the encodings are somewhat arbitrary. For 
example, we had some discussion about how to encode lists in RDF 
triples format, and chose one technique from the  several options on 
the table.

The requirement that DAML+OIL be encoded in RDF has therefore 
required that the parts of DAML+OIL syntax that do not naturally fit 
into RDF triples are encoded as structures built from RDF triples. 
The result is that the relationship between DAML+OIL syntax and RDF 
syntax is not simply an inclusion of one syntax in another (the kind 
of thing that in the simplest case would be one language's BNF being 
simply an extension of the BNF of the other language, in the same 
kind of way that predicate logic is a syntactic extension of 
propositional logic) but instead is more like an implementation of 
(part of) one syntax in another, rather like the relationship between 
Prolog and LISP when a Prolog interpreter is implemented in LISP. The 
set of RDF triples into which a piece of DAML+OIL is transcribed 
therefore has two rather different parts. Some of it has the same 
meaning in RDF as it has in DAML+OIL (well, not strictly, but close 
enough that we could make it the same), but other parts of the 
resulting RDF are not about the DAML+OIL semantics at all, but 
instead are about the *syntax* of the DAML+OIL. And since those 
encodings are arbitrary, there is no principled way to define away 
the RDF triples that describe them. This means that, as Peter has 
explained, there is no way to extend an RDF model theory to a 
DAML+OIL model theory in a way that preserves both the encoding of 
DAML+OIL in RDF triples and also, simultaneously, the intended 
meanings of both the RDF triples and the DAML+OIL.

This has been kind of obvious from the beginning of the DAML project, 
right? (Why did we even get involved in defining list structures, if 
we could have just used RDF triples as our syntax?)

There are several ways to try to fix this, if it seen as being a 
problem. One might be to redefine the relationship between the model 
theories so as to account for the syntactic encodings. This however 
would be a research problem. (I started looking at this, and I think 
it would require some quite serious changes to the RDF model theory 
in any case, as it would need to support recursion on the DAML list 
constructs. )  A quicker and more direct way to go, which Peter is 
suggesting, would be to simply abandon the requirement that all of 
DAML+OIL syntax be encoded in RDF triples, and allow DAML+OIL syntax 
to be a genuine extension of RDF syntax. Some of DAML+OIL consists of 
RDF triples, but some of it really does not. This has the merit of 
simplicity and semantic clarity, and it also makes the relationship 
between RDF and DAML+OIL meanings clearer; but it kind of demotes RDF 
from its position in the center of the SW universe, which may be 
against the teachings of some church or other.

Pat

-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Wednesday, 12 December 2001 07:34:42 UTC