Re: SUO: ontology as science
I don't know if I'm on the black list of monolithist or not. But I really feel
that this is all mixing different things:
-one versus many upper-level ontologies
-'single module' versus modularity
-upper-ontology and reasoning context
It seems to me that most people who have argued for a single ontology have not
rejected modularity for as much. It has been said several time that the
treatment of modularity would find a nice source of inspiration in the work of
Guha and Lenat. It has also been said that Cyc is using only one upper-level
ontology.
Cyc uses microtheories for a variety of different purposes, most importantly:
-compartimenting knowledge (cause you need small KBs even when you have a large
one)
-allow variant theories (or phrase it: multiple ontologies)
-allow reasoning context and picking out the relevant bodies of knowledge for
certain queries
(See also DL on conceptual spaces. Amanda Vizedom is a reference here, she
might have some publication available.)
IMHU, there is no organic link between modularity and plurality (and their
opposites). But moreover, bringing reasoning contexts when speaking of
constructing an upper-level ontology is just plain irrelevant.
I'll readicalize my view a bit so that the message is plain. Modularity,
reasoning contexts and all those nice things are implementaion issues. The
issue for the SUO is to build a theory. You don't build a theory by sticking
bits of axioms and principles one with the other because one or another of your
acquaintances likes those bits.
If it proves that there are many variants that are required according to this
group, then we first need to lay down all the theories. Then we may worry about
how to interrelate them (which are inconsistent, extensions of other, are there
common fragment, and so on).
Then you can worry about the cosmetics and how you will structure the relevant
modules or microtheories -- theoretically, the mere idea of a microtheretic
system, as shown by your lattice of theories theory, is completely trivial.
Best
Pierre
>
> Bill and Matthew,
>
> We are in agreement, but there are still some monolithic fans
> among us who need more convincing.
>
> MW>I agree with you [WB].
> >
> >I have actually suggested a couple of times that the important thing
> >we need to do is establish a process, rather than define a deliverable,
> >but it often takes several attempts to make a point.
>
> WB> Yet this characteristic/feature - semantic drift
> > - requires some method be included with the SUO to govern its
> > evolution over time in response to changing requirements.
> > (Likewise, some method is needed to account for language and
> > contextual differences.)
>
> I wholeheartedly and enthusiastically agree.
>
> See my previous response to JA and JF in this same thread.
> Cyc supports contextual differences with microtheories, and
> we need a similar modular mechanism in any proposed SUO.
>
> John Sowa
>
--
Pierre Grenon
IFOMIS Uni Leipzig
Haertelstr. 16-18
04107 Leipzig
http://people.ifomis.uni-leipzig.de/pierre.grenon/
pgrenon@ifomis.uni-leipzig.de
phone: 49(0)351971672
fax: 49(0)3519716179