Re: clarifying clarifying ontologies
sowa@west.poly.edu (John F. Sowa)
Date: Tue, 8 Aug 1995 13:52:44 +0500
From: sowa@west.poly.edu (John F. Sowa)
Message-id: <9508081752.AA11948@west>
To: cg@cs.umn.edu, forbus@ils.nwu.edu, hovy@isi.edu, phayes@cs.uiuc.edu
Subject: Re: clarifying clarifying ontologies
Cc: doug@csi.uottawa.ca, fritz@rodin.wustl.edu, srkb@cs.umbc.edu
Content-Length: 1519
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk
I would like to add my $0.02 as an an enthusiastic augmentation to
the $0.02 that Ken Forbus threw into the pot. I very much agree with
his point about analying predicates and using a compositional or
combinatorial method of generating the taxonomy, which is much less
fundamental than the basic distinctions that come out of the conceptual
analysis of predicates.
In my paper for the IJCAI workshop on ontology, I emphasize distinctions,
combinations, and constraints:
"The distinctions generate differentiae, the differentiae generate
categories, and the axioms (or constraints) preserve possible
categories and filter out the impossible ones. Axioms associated with
the categories are inherited through the hierarchy to subcategories.
They form the nuclei of the microtheories for specific domains."
In my paper, I give some examples of pretty trees and lattices for displaying
the categories, but I emphasize that they are primarily a visual aid, rather
than a fundamental principle. The distinctions, combinations, and constraints
(a.k.a. axioms) are fundamental, and the trees or lattices are a useful, but
secondary byproduct.
And for the record, I follow the medieval tradition of using the word
"predicate" as a translation of Aristotle's word "category". Both words
are intensional terms; the set of things that satisfy a predicate are
its extension. But the extension is always relative to the current
state of the real world or hypothetical world under consideration.
John Sowa