Re: 11179-2 Classification
Fritz Lehmann <flehmann@orion.oac.uci.edu>
Date: Tue, 1 Aug 1995 17:01:57 -0700
From: Fritz Lehmann <flehmann@orion.oac.uci.edu>
Message-id: <199508020001.AA20377@orion.oac.uci.edu>
To: bsr@utu.premenos.com
Subject: Re: 11179-2 Classification
Cc: bsr@premenos.com, cg@umn.edu, edi-new@tegsun.harvard.edu,
fritz@rodin.wustl.edu, srkb@cs.umbc.edu
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk
In the EDI Basic Semantic Repository (bsr) list
M. Paul Hawes wrote, on the ontological "object taxonomy"
that was to be in the ISO 11179 standard for data semantics,
---begin quote---
I was in Brazil when we made this decision at an SC14 meeting. I am
unaware of any "dispute" -- we just decided there was not going to be a
single classification scheme to satisfy everyone's needs. We decided to
describe how and why to create a classification scheme. An example will
be in an informative annex. Bruce Bargmeyer has become a co-editor along
with the former editor Bill Kenworthey. Both are from ANSI X3L8.
---end quote---
This is reassuring if the informative annex is still going to be a major
thrust of ISO 11179, even though it confirms that the ontological
taxonomy has, for now, been excluded from the main part of the 11179 standard.
My concern was not to defend the taxonomy as-is (in fact my original question
was about the best place to suggest major changes), but rather to defend the
idea that _some_ (set of) agreed-upon concept (or 'basic semantic unit')
systems for commercial and real-world conceptual building-blocks is needed
in order to automate semantic integration of different data and knowledge
bases -- and to integrate electronic-commerce/EDI messages with those systems.
Thus it seems important to know to what degree there will still be a
"common ground taxonomy" as an important component of ISO 11179,
and to what degree the EDI BSR is harmonious with, or diverges from, that
taxonomy. I can imagine perfectly good reasons why they might diverge,
but if they do diverge I hope the reasons for diverging are in fact good.
The business of ontology/taxonomy building is complicated, subtle
and controversial. I was pleased to see that someone on X3L8 had considered
the important Pangloss Ontology Base. The 'data semantics' world is
only a small part of the efforts worldwide to build practical semantic
taxonomies. My lecture at the Ontologies workshop at IJCAI-95 in Montreal
in a few weeks will partly deal with how all these systems might be be
linked to each other; I have considered the 11179 "object taxonomy" and
Denis Hill's "Basic Semantic Repository" as candidates for integration
with the rest. One of the real advantages of linking-up is that you
can exploit the thoughtful prior work of others (in areas where their
expertise might exceed your own) while still maintaining your automonomy
to set things up the way you want in the areas you care about.
The "how to create" parts of ISO 11179 seem very well thought out,
but in my view the biggest challenge is still in the _content_. In
another standards effort, the X3T2 Conceptual Schemas Modeling Facility
(CSMF), there were also two parts: the (much discussed) issue of the
representation languages and "how to create", and the (far less discussed)
issue of what the common semantic content should actually be. I have
not yet heard what the recent X3T2 meeting on the CSMF did about the second
issue, but a member of that group indicated earlier that [as with ISO 11179],
it had seemed easier to confer about methodology than about actual content.
Top people in the Artificial Intelligence field see the same problem there.
So I've been warning about this tendency (to focus on methodology over
ontological/semantic content) in response to a possibly-widespread syndrome in
data standards which might ultimately defeat the main goal of
enabling the successful integration of different systems.
Yours truly, Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel: +1(714)-856-0671 email: fritz@rodin.wustl.edu
=============================================================