KR Standards meeting trip report
Tom Gruber <Gruber@sumex-aim.stanford.edu>
Message-id: <2855950113-10832308@KSL-Mac-69>
Date: Mon, 2 Jul 90 16:28:33 PDT
From: Tom Gruber <Gruber@sumex-aim.stanford.edu>
To: Summer Ontology Project <shared-kr@sumex-aim.stanford.edu>
Cc: Edward Feigenbaum <eaf@sumex-aim.stanford.edu>,
Bob Engelmore <rse@sumex-aim.stanford.edu>,
Steve Cross <cross@darpa.mil>,
Peter Friedland <friedland@pluto.arc.nasa.gov>
Subject: KR Standards meeting trip report
Last thursday and friday the advisory board on knowledge representation
standards met at Bob Kahn's NRI offices. DARPA has funded a two year
plan, managed by ISI, to promote the development of requirements for
knowledge representation standards and related means to foster shared,
reusable knowledge bases and knowledge-based systems and tools. The
plan is essentially to hold a series of meetings and workshops, write
reports, etc. The emphasis of the effort is broader than language
standards (analogous to Common Lisp). In the short term the
"standardization effort" is also interested in developing *requirements*
for tools and methodologies, and *recommendations* for research that
needs to be done to facilitate knowledge sharing and reuse. The work is
being accomplished by working groups specializing in: interlingua
(interchange format like KIF), common representation systems
(implementations like Prolog and KEE), external interfaces (query
interfaces to databases, client/server protocols, etc), and shared
resuable knowledge bases (ontologies, generic tools, etc). Marty
Tenenbaum and I chair this last group.
At the meeting there was plenty of talk about the goals of the effort,
visions, status reports from working group chairmen, etc. From the
discussions it is clear that the summer ontology project is a valuable
experiment for the whole KR standards effort. Here's some ways in which
it can contribute.
* The example ontology and representation examples are critical feedback
for the folks trying to hash out a low-level interlingua.
* The experience of trying to build group ontologies generates valuable
data for designing more extensive experiments. The sponsors want to see
similar efforts by other groups in other domain areas, and other groups
using the results of our work, and in the near future.
* Our strategy of trying to use LOOM to do term maintenance and Epikit
to represent the "assertional" knowledge will help inform the debate on
what information is needed for KB development in addition to KB
interchange.
* Experience with our method of evaluation -- model evaluation via
validation questions -- and the actual corpus of query/explanation pairs
will be valuable for the specification of evaluation/validation methods
for shared ontologies. Part of the plan for the standards effort is to
develop test suites by which people can determine whether they are
compliant. As you can imagine, it is a research problem to figure out
how to do this for representations and ontologies. That's some ground
we will be breaking.
One of the "action items" for me coming out of the this meeting is to
organize the official working group on shared reusable knowledge bases.
The charter of this group is to draft specifications for shared reusable
ontologies, KBs, and tools. The first thing to do is come up with some
scenarios of use for a "vision statement" -- that is, specific visions
of how the world would be a better place if shared KB technology was in
practice. If you are interested in actively participating in this
working group, let me know. A first drafty spec is planned for October,
so that means getting starting now.
--tom