DAML for syntax [Strawman DAML+OIL Query...]

From: Dan Connolly (connolly@w3.org)
Date: 09/07/01

pat hayes wrote:
> >Richard Fikes wrote:
> > >
> > > As I mentioned yesterday in the telecon, a student of mine, Yulin Li,
> > > and I have designed a simple language for querying DAML+OIL knowledge
> > > bases.  The language is specified as a DAML+OIL ontology so that both
> > > queries and the results obtained from asking a query are represented in
> > > DAML+OIL.
> >
> >I appreciate the effort of specifying the language as a DAML+OIL
> >ontology.
> Well, I wonder what the point of this effort was, and would like to
> raise this as an issue for discussion.

Very well... but the discussion isn't really about the query
language proposal, so I've edited the subject header field...

> It seems to me to be:
> (a) completely pointless, in the strict sense that it provides no
> useful functionality or understanding of the language (ie the query
> language) to describe its syntax in DAML+OIL as opposed to, say, BNF.
> All that this enables a hypothetical DAML reasoner to do is to parse
> the expressions of the query language. Using an ontology language for
> parsing seems a very poor design decision; at the least, one that
> should be discussed on its merits rather than simply assumed to be
> somehow a Good Thing.

I agree that there isn't any particular point to be made
in using current DAML+OIL as opposed to BNF; but there is a point
to be made by using *some* formal description of the syntax
in addition to the examples and some prose.
DAML is the formalization we're working on, and I appreciate
the effort of trying to apply it to the task at hand.
For example, I'm learning something from the points you raise below...

We could go back and formally specify syntax and semantics for
some BNF, with all the hairy internationalization and
other interoperability issues, but that would seem
to be quite a distraction.

> (b) actively misleading, in the sense that it suggests that the
> purpose of DAML+OIL is to be a syntax specification language, which
> as far as I was aware wasn't ever even close to the intended goal of
> the project;

DAML is an effort to build an agent markup language; surely
one of the things agents will need to talk to each other about
is syntax. Perhaps applying the 2001-03 DAML+OIL language
to the problem of syntax description is a stretch; but
what better way to figure out what DAML needs next than
to apply the current spec to the problems at hand and
see what comes up?

> (c) based on a basic misapprehension about the nature of descriptive
> languages, in that part of the very idea of a *syntactic*
> specification is that it describes domains of recursively defined
> finite entities to which results such as the second recursion theorem
> apply, whereas descriptive (assertional) languages like DAML+OIL (and
> RDF and FOL) have an extensional semantics which is (because of
> Goedel incompleteness) inherently unable to fully capture the notion
> of finiteness or recursion.

I'm afraid I'm guilty of this misapprehension. You have
raised this point about the inability of FOPL to express
finiteness/recursion a number of times now, and I have
attempted to research it to the point where I understand,
but I'm not there yet.

If the 2001-03 DAML+OIL thingy can't express finiteness
nor recursion, I'm interested in figuring out what
we could layer on top to give DAML the ability to express it.

> >We chuckled on the phone about what an obscure mechanism it is
> >for communication in the group...
> >but I hope each of us has at his/her disposal some tools
> >that render DAML+OIL intelligible to us in some familiar
> >notation/interface.
> I'm afraid I do not have access to any that I know how to use.  I
> wish there was one, indeed. Do you know of any?

As I said in a subsequent message, I find rendering DAML
as RDF/n3 helps me understand it; there's an online
conversion service...

	(it converts in both directions)

I've implemented the basics of reading/writing RDF
in KIF syntax in a way that could be integrated into this
online service, if you're interested; (the source
is also available, and I could perhaps help you
install it; I think there are some issues with
the python XML tools and the mac platform, though).

As Danbri pointed out in his follow-up

  The FRODO RDFSViz Tool

is available...
but I'm not sure it pays attention to RDF vocabularies
other than RDFS, so I'm not sure how applicable to
DAML it is.

I think protoge, oiled, etc. are also available,
though I don't have first-hand experience with them.

Dan Connolly, W3C http://www.w3.org/People/Connolly/

This archive was generated by hypermail 2.1.4 : 04/02/02 EST