On Jan 25, 2007, at 11:43 , Steve Ray wrote:
This has been very interesting, but I must point out that the original goal was
not to come up with a definition of ontology.
We were trying to come up with a categorization scheme into which we could place
everything that anyone calls an ontology, whether it is computer processable or
not, consistent or not, correct or not.
As part of such a categorization scheme, I agree that it is useful to have good
definitions to distinguish things on this spectrum, and the current discussion
is addressing that for things at one end of the spectrum. We might well come up
with a definition that this community agrees is a "true ontology", but the point
is that lots of people are using the word in ways that we may or may not agree
with. Rather than simply insisting that they are wrong, I believe it is more
productive to clarify how these various animals are different.
It strikes me that there is a fundamental conflict (or at least a difficulty) here with your stated goals:
* Anything anyone wants to call "(an) ontology" must receive due consideration. Call this class of things O
* There is a coherent classification scheme by which things in O may be usefully described and differentiated
I just don't see how this is possible (or for that matter useful) without some minimal requirements. Here are some suggestions:
(1) The things in O are artifacts to be used to guide or drive the operation of software that performs some intended functions. It might be useful to talk about what those functions are
(2) Anything that can meets requirement (1) must be expressible in some logical system. Else it's hard to say by virtue of what it meets requirement (1). Seems the Church-Turing Thesis is relevant here, but I don't have time to work out the ramifications.
(3) Anything that meets requirement (2) will therefore have a formal syntax and semantics that govern the interpretation of the terms in languages expressed in the given logical system. Two things are relevant here - one is a theory of language (we want to talk about concrete and abstract linguistic expressions of things in O) and the other is semantics (we want to talk about the meaning of things in O). Relevant here would be things like Chris Menzel's work on Ontology Theory.
The above are not unproblematic. Requirements 1-3, for example, are satisfied by any arbitrary computer program.
So.. All this leads to a fundamental difficulty. If you want to let everything in that someone wants to call "ontology" you're going to have a huge (and IMHO intractable) problem on your hands, and one that is not likely to be solved by committee. The other way to go is the way of philosophy, and as Chris and Pat Hayes have pointed out, that's going to lead to another black hole.