Re: Common ontologies

• Mail folder: SRKB Mail
• Next message: Sam Hunting: "Montreal Conferene"
• Previous message: Patrick Cassidy: "Common ontologies"
• Maybe in reply to: Patrick Cassidy: "Common ontologies"

Date: Sat, 10 Jun 1995 00:47:37 -0700
From: Fritz Lehmann <flehmann@orion.oac.uci.edu>
Message-id: <199506100747.AA06928@orion.oac.uci.edu>
To: cg@cs.umn.edu, srkb@cs.umbc.edu
Subject: Re: Common ontologies
Cc: pdoudna@aol.com
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

     Pat Cassidy's proposal for ontology sharing is a good sign,
and it will be interesting to see if the other "players" come
forward with proposed  conceptual agreements by the time of the
Montreal workshop.  However, he included:
---begin quote---
    (4) The resulting set of possible "common" ontologies can form
    the starting point for the discussion of the feasibility of finding
    a single common system that a majority can agree on.
---end quote---

    I suspect that "finding a single common system that a majority can agree
on" is a premature (and maybe inherently impossible) goal, but that a related
goal _is_ feasible: establish what Kevin Knight called "anchor points"
at which major systems are tied together.  In negotiating and hashing out
the proper anchor points, I would expect people to have new insights on
what their own categories are intended to mean.  Exact match may be
impossible (between a concept in one system and a corresponding concept
in another) as I will opine in my Montreal talk, but useful, valid
mappings are feasible.  "PHYSICAL-MATERIAL" in Pangloss might not quite
mean the same thing as "STUFF" in Cyc, or whatever class is closest, 
but a fairly reliable mapping can nonetheless be made.  If so, one system
may be able to exploit some inferential capability, or natural language
information, in the other.  If some derivative of RatNoCo-Roget class
316 "Materiality" is similarly mapped, Roget-based inferences and word
lists will similarly become available.  Etc.  This is not as hard as
the actual combining of upper models (as was done by hand to Penman, Ontos
and Wordnet for the Pangloss Ontology Base, for example).  I cannot imagine
the prima donnas of this field doing much to alter their upper hierarchies
to correspond to others, nor should they since these hierarchies have
different purposes.  I'd aim only for translation and cross-linkages, not
the unification suggested by Cassidy's point (4).

                          Yours truly,   Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-856-0671               email: fritz@rodin.wustl.edu
=============================================================