Re: Some references about ontology and analogy: SUO redux
John,
About your challenge proposal: it seems perhaps too advanced, unless the
analysis of scenes has progressed beyond what I think it has. If you used
instead a geometric representation of scenes (as in a flight simulator) and then
asked the computer questions about its internal representation, that would still
be a challenge, and if successful could be extended to the scene recognition
problem. This would be comparable to using electronic text to test language
understanding programs, rather than requiring that they listen to a voice. The
perception element could come from extension at either end (pictures or voice)
but I don't know if it's realistic to hope that picture analysis will get
advanced enough, even in 20 years.
> . . . after spending $70
> million on putting some very good minds to work on the problem,
> the Cyc project still doesn't have a version we would be happy
> to standardize.
>
I don't know why Cycorp cannot devise an ontology more appealing to more
people, but Cyc appears internally sound to me, from what they have made public.
It also appears unnecessarily complex, especially in its overuse of
metaclasses, and stuff like that makes an already complex system more complex
and incomprehensible to all but the most determined users. I suspect that they
made design decisions at an early phase and became increasingly reluctant to
change the upper levels. The extensive use of microtheories may well have
allowed them to ignore fundamental problems by pretending that they could cut
off parts of the ontology from each other. There's nothing wrong with
microtheories when there are real logical incompatibilities, real contextual
issues to address, or real logical isolation, but it may also just be a hurried
ontologist's way of ignoring problems. I have never seen the full Cyc system,
and cannot judge how they are used.
>
> PC> I just want to get on with the task of building the best
> > upper ontology that the best minds can devise in the present
> > state of our art, keep modifying it as experience reveals
> > its inadequacies, and in the interim use it as a tool that
> > will help us make progress on the other issues that you
> > have eloquently pointed out, by improving the reusability
> > of the results of our investigations.
>
> My main concern is with the question of what to standardize.
> The best minds of the past two millennium haven't reached an
> agreement on a good upper ontology, and after spending $70
> million on putting some very good minds to work on the problem,
> the Cyc project still doesn't have a version we would be happy
> to standardize.
>
I don't expect everyone to agree with any product that is agreed to by a
committee, even one with representatives of over 40 different groups involved in
its development, and even if they were on the committee. But if as few as 30
percent of those who use ***some*** upper ontology decided to create their
domain ontologies by reference to the proposed common upper ontology, we would
have a working standard and would be way, way ahead of where we are now. I
think that there are plenty of good reasons for thoughtful people [I include
myself ;-)] to object to every standard that has been proposed thus far, and
none of us have had adequate motivation to change whatever our favorite
representation is. Motivation could come from two sources: (1) an impressive
demonstration of the utility of some proposed standard; or (2) adoption by a
significant and diverse community of practitioners, even before an eye-opening
knock-your-socks-off demonstration program could be developed.
I think that the latter will be easier to achieve (with the 30% adoption
criterion) than the former, but if anyone has a good idea of a demo program that
would be really impressive, I would be willing to help build one to the extent
that my skills are relevant and could find the time. I think that a
Natural-Language program would be a good demo, but it seems hard to imagine that
a volunteer group could improve on what groups of ten to 30 people are already
doing. A few NL groups are already using ontologies. If any groups building
question-answering programs would like to collaborate with ontologists to try to
improve their systems, that project might serve to focus our efforts into a
channel that could prove what most of us feel about the utility of ontologies.
> The second note, from Jim Schoening, suggests that we
>
> Demonstrate how a hierarchy of a single upper, multiple
> domain, and multiple sub-domain ontologies will enable some
> useful level of queries and inferencing across domains.
>
This would be helpful, if the problem were recognizably non-trivial. Still,
I fear that a demonstration of cross-domain knowledge retrieval would only be
impressive to someone who has already struggled with the problem. I suspect
most casual observers will not know enough to be impressed by what might look
like just another database query system. If there is someone who has a sticky
cross-domain information retrieval problem, and who, suitably motivated, would
be in a position to put some serious money into ontology development, that would
be a good place to try a demo, by our community or any other capable group.
People with funds, bring us your worst interoperability problems, and let us
show what we can do. Paupers need not apply.
> In it, I make the following suggestion:
>
> ... that we address the problem of working on a document
> about principles for developing and defining ontologies.
>
I would worry about the loss of time developing agreement on general
principles of ontology development; that's what Guarino et al have been working
on for years. But if we were to start our upper ontology project by first
developing agreement on the principles **we** will use to develop **that
specific ontology**, this would, I believe, be worthwhile, provided that we
agree to finish this preliminary phase in no more than two months tops, and then
get on to the construction phase.
Pat
--
=============================================
Patrick Cassidy
MICRA, Inc. || (908) 561-3416
735 Belvidere Ave. || (908) 668-5252 (if no answer above)
Plainfield, NJ 07062-2054
internet: cassidy@micra.com
=============================================