RE: SUO: RE: Re: Abstraction, Analogy, Example, Icon, Metaphor, Model, Morphism, Paradigm, Prototype, Simulation

[apease <apease@teknowledge.com>]

Philip,
   Thanks for the clarification.  On both your points, I could give the 
example of the DARPA HPKB project in which we used an ontology of some 
400,000 axioms and 40,000 terms (Cyc).  We had three different test 
applications we created in quite different domains (international politics 
and trade, army battle planning and battlefield engineering and 
construction).  I don't believe there was any case of major reorganization 
of the upper ontology, nor any case where it was difficult to add new terms 
to the ontology.  I should say that there were concepts that were difficult 
to formalize because they were inherently complex, but it wasn't a problem 
with the parent ontology overly constraining the addition of new content.

Adam

At 10:11 PM 10/31/2000 -0500, Philip Jackson wrote:
>Adam,
>
> > Phil Jackson wrote:
>
> > >I like this quote because it suggests that to imagine we must
> > >know what we
> > >do not know, and then consider ways to learn what we do not know.
> > >Imagination in the form of thought experiments, and metaphorical
> > >explanations, is vital to the extension of human knowledge.
> > >
> > >SUO has a choice whether to develop a conventional, formal,
> > >logical ontology
> > >that captures knowledge, yet which may be brittle, and/or to develop a
> > >"natural, semiotic, linguistic, ..." ontology that is extensible,
> > >metaphorical, able to represent what is not known, etc. The
> > >latter approach
> > >is much more difficult and involves much more research, yet may
> > >be stronger
> > >in the long run. There is of course no reason why both tracks cannot be
> > >pursued in parallel, with some benefit from cross-pollination of
> > >ideas, nor is there any reason why either approach cannot produce a
> > >worthwhile standard upper ontology.
> >
> > Why do you believe that a formal ontology is brittle, and how would you
> > characterize a "linguistic" ontology?
>
>I think a formal ontology may be brittle, not that it necessarily is.
>
>I think the brittleness might be experienced in various ways, e.g. 1) as the
>ontology becomes larger, people may find it is sometimes difficult to add
>new concepts into the ontology without restructuring it, or to integrate the
>ontology with other large, independently developed ontologies; 2)
>ultimately, if we build intelligent systems using the ontology, there may be
>some concepts that the system itself cannot easily add to its ontology.
>
>Thus, we may suppose that humans could in principle always add new concepts
>into an ontology, yet ask whether in principle a system could always use its
>ontology to add concepts for new domains into itself.
>
>However, I am open to logic and evidence that a conventional formal ontology
>can be very large and yet not brittle; I recognize that much of the
>motivation for ontology research is to avoid the brittleness encountered
>with limited, domain-specific systems. And so I do not see this potential
>for brittleness as a reason not to undertake creating an SUO -- rather, to
>the extent that a very general SUO is created, it should help avoid
>brittleness.
>
>Beyond the characterizations I've given above, I'd say that to the extent
>that an ontology (and the conceptual language that supports it) provides
>very flexible, general mechanisms for extending its concepts into new
>domains using concepts that it already contains, then it should be less
>brittle. It seems that both the initial concepts, and the linguistic
>framework for expressing new concepts, could benefit in generality and
>flexibility from research in linguistics, semantics, semiotics, and other
>fields. I envision that such a system would be able to use metaphors to
>represent concepts in new domains; represent concepts that are ambiguous, or
>associative, or prototypical and irregular, as well as formal and regular;
>create meta concepts to represent knowledge about its current concepts,
>including the limitations of its current concepts; etc.
>
>Such an approach may be viewed more as an extension of a conventional purely
>formal ontology rather than as an alternative. Certainly, many people have
>done and are doing research on topics like representation of metaphors,
>irregular concepts, etc. And granted, any such system would still ultimately
>be described as a formal system, since it could be written in a formal
>programming language or specified in FOL. Essentially such a system would
>provide a formal representation of conceptual mechanisms often considered
>informal.
>
>Phil Jackson
>-------------------------------------------------------------------
>"Imagination is more important than knowledge. Knowledge is
>limited. Imagination encircles the world." - Einstein
>-------------------------------------------------------------------
>Standard Disclaimers. www.philjackson.prohosting.com

-----------------
Adam Pease
Teknowledge
(650) 424-0500 x571