Re: SUO: SUO-ML and SUO Namespaces: Towards breaking the monolith ...
I support Leo's suggestion (below)
I believe that a subsumption lattice of ontology theories
would be the best thing that could come in the long run from
the SUO effort (and all the other ontology efforts). I think
that is what Leo believes. Given that, the big question is
how to get there, especially without major funding. I think
Leo's suggestion is a realistic way to make some progress
that will not be wasted when the optimum state is achieved.
My focus in earlier postings was to explore how we could
minimize the number and diversity of ontological theories.
There will no doubt be real differences in approach that force
different groups to adopt inconsistent ontologies, but I think
that most of the differences one finds now are due to divergence in
approach by different groups that is caused simply by the absence
of single widely-accepted upper ontology. I think that,
given an assumption that there will be a lattice of theories, the
focus of a "standards' group is appropriately directed at
isolating and minimizing the differences to those irreducible
few that users won't give up in exchange for complete
interoperability. Toward that end, if we try to build a
subsumption lattice of existing theories, it could force us
to determine precisely where the essential contradictions between
theories occur, and can help to isolate the differences into
small modules that are genuinely inconsistent.
I sincerely hope that there is at least some minimum at
the highest "object" level of the subsumption lattice that
can be agreed on by all ontologies, to form the "base
ontology" shared by all. This is another worthy focus for
a "standards" group. John Sowa has often emphasized the
essential role of the "distinctions" in an ontology --
I also agree with that, provided that the relations and
functions (with their defining axioms) are considered as
"distinctions" that define the differences between
concepts. For this reason, I would hope that as many as
possible of the distinctions, relations, functions,
would be included in the common "base ontology". Different
groups may use such distinctions to define inconsistent
theories, but we would be in a better position to
understand the differences in such a case.
I do not yet have a Unix system available, so I am now
ordering the materials to install a Linux system so that
I can install the OpenCyc platform. Until then I will not
be able to comment on OpenCyc.
Pat
=============================
Leo Obrst wrote:
> ...or, What to do until the future arrives (IFF, Theory-lattices)
>
> Per my last message, here's a very simple, low-tech proposal: why don't
> we define a set of ontology content metadata, called SUO-ML (for
> SUO-XML, for example), each term of which can define a SUO-NS
> (namespace). These content terms will be ontology subject areas, along
> the lines of: Space-Time, Part-Whole, etc. Ontology builders, if they
> follow the SUO-ML standard, will build their ontologies and identify
> them by SUO-NSs. These content metadata terms will act like the usual
> metadata terms, i.e., they are interpreted mainly (at least initially,
> but eventually not just) by human beings. Then ontologies can be found
> across the Web by searching on SUO-NS<area>, recognizing that there are
> going to be some inconsistencies. When you build an ontology module, if
> you follow the standard, you "tag" your module with the label. This
> approach will help us at least work toward the problem Jim F. has (as do
> we all), and towards modularity and non-monolithicity. Labels are
> notoriously short on semantics, but at least they have some, to a human
> being. If a given SUO-ML category also had a natural language
> description (and of course your ontology under that label also had such
> a NL description), at least human beings would get some of your meaning.
>
> The metadata could even be structured into hierarchies, if desired
> (example: under Space-Time, two branches could be 3D and 4D). Over time
> these could become fairly elaborated (a classification of human
> knowledge? uh, I won't be so grandiose). And eventually when the IFF (or
> comparable facility) matures, we can more easily make the transition to
> a lattice of theories.
=============================================
Patrick Cassidy
MICRA, Inc. || (908) 561-3416
735 Belvidere Ave. || (908) 668-5252 (if no answer)
Plainfield, NJ 07062-2054 || (908) 668-5904 (fax)
internet: cassidy@micra.com
=============================================