Re: Classification Systems? (Fritz Lehmann)
Date: Wed, 31 Aug 94 11:45:32 CDT
From: (Fritz Lehmann)
Message-id: <>
Subject: Re: Classification Systems?
Precedence: bulk

Bill Rich of IBM asked:
------begin quote----------
Question:  Who knows about classification systems?  We are interested in
the theory behind classification systems, existing classification
systems, and the history of classification systems.

Background:  As part of a project we're working on, we've begun to
classify objects and relationships between objects.  Classification
systems exist such as Library of Congress classification, Dewey Decimal,
Binomial nomenclature, but is there a science to the classification?  Are
there any theories of coherency, consistency, and completeness?  What are
the classification systems that are generally accepted?

Thanks.  Bill --
--------end quote-----------

     Yes, there is a science of classification systems.  In
fact there are several sciences, which differ in emphasis.
For "ontological" classification, including the theory
behind various library classification systems, try joining
ISKO, the International Society for Knowledge Organization,
Woogstrasse 36a, Frankfurt D-60431, Germany.  There is
also "numerical taxonomy" and statistical classification;
for this, try joining the Classification Society of North
America, which publishes the Journal of Classification.
There are two schools of lattice-based classification,
that of Rudolf Wille at Technische Hochschule Darmstadt
and that of Yu. Shreider in Russia (the latter often
publishes in English translation in the
journal Automatic Documentation and Mathematical
Linguistics).  Strict classification of descriptions
based on necessary and sufficient conditions is
the hallmark of some KL-ONE systems in AI.  If merely
necessary conditions are present, a prototype-based
approach to classification is necessary, as described
in Franconi et al.'s article in my "Semantic Networks
in Artificial Intelligence" collection (Pergamon Press,
Oxford, 1992, and Computers & Math. with Applications,
v. 23, no.s 2-9, 1992).

     Dewey Decimal and Library of Congress are dominant
in libraries only because they are the basis for free
book classification services.  Theoretically they are
obsolete compared with the COLON system used in some
libraries in India.  The latter is "faceted" into different
aspects like Matter, Energy, Space, Time.  The existence
of such "ontological dimensions" (facets) is an exciting
area for research.  It affects the abstract algebraic
structure of the type hierarchy, as my recent papers have
sought to emphasize.

     Broadly speaking, there have been hundreds of classification
systems, general and specific.  See the list of 158 concept-
systems in my CCAT report in the Proceedings of the Peirce
Workshop held at the Conference on Conceptual Structures this
month (ICCS-94).  Some systems have formal theories of classification,
some don't.

     The more thoughtful theorists in the area of Knowledge
Acquisition/Engineering have developed principles for
classification, as have some object-oriented design gurus.

                          Yours truly,   Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-733-0566  Fax:(714)-733-0506