Update on ANSI and ISO activities
sowa@watson.ibm.com
Date: Tue, 17 Nov 1992 13:11:52 -0800
Message-id: <199211172108.AA12836@venera.isi.edu>
Comment: SRKB Distribution List
Originator: srkb-list@isi.edu
Errors-To: neches@ISI.EDU
Reply-To: <sowa@watson.ibm.com>
Sender: srkb-list@ISI.EDU
Version: 5.5 -- Copyright (c) 1991/92, Anastasios Kotsikonas
From: sowa@watson.ibm.com
To: Multiple recipients of list <srkb-list@ISI.EDU>
Subject: Update on ANSI and ISO activities
During the past few months, there have been several developments that
are significant for the Knowledge Sharing Effort and the conceptual
schema work in ANSI and ISO.
Two weeks ago, while I was in the Bay Area, I visited Stanford to talk
with John McCarthy and Mike Genesereth about contexts. We had a fairly
lengthy discussion that made good progress. It looks as if we should be
able to use the definitional mechanisms in KIF to define the kinds of
contexts that I have been using in conceptual graphs and to define the
lifting rules that McCarthy has been developing in his theory of contexts.
Following is a brief summary of the approach that Genesereth and I have
proposed. McCarthy still has a wait-and-see attitude until we get the
details formalized.
1. McCarthy's basic predicate for contexts is ist(p,c), which says
that proposition p is true in context c. He then defines a number
of "lifting rules" for importing and exporting propositions into
and out of contexts under various conditions.
2. My basic notation for contexts in conceptual graphs is a box that
contains a set of other conceptual graphs. Relations attached to
the box specify how those graphs are to be interpreted: as literals
that are being quoted; as propositions that are being asserted; or
as a theory (or microtheory) that says something about some aspect
of the world. Then outside the box, other conceptual graphs can
make metalanguage statements about the way the graphs inside the
box interact with one another. Context boxes and metalanguage
statements may be nested to any depth.
3. Since the CG context box is nothing more than a simple enclosure,
it can be simulated in KIF by the usual QUOTE operator (or the
backwards quote, which allows variables inside the box to be bound
by quantifiers outside). The relations attached to the CG box can
be represented by KIF predicates that say something about the quoted
structures. Then the metalanguage CGs outside the context box can
be mapped to KIF metalanguage statements. With such a mapping, it
appears that the context facilities that I have been defining for
conceptual graphs can also be defined in terms of the primitives that
are already present in KIF.
4. The next task is to formalize the details of that mapping and to
test it out by seeing whether we can use it to represent the lifting
rules that McCarthy has proposed and that Guha has implemented in Cyc.
McCarthy gave us a copy of his "Notes on Formalizing Contexts,"
which we can use as a test case.
This discussion does not mean that KIF and CGs both subsume McCarthy's
theory of contexts. Rather, it means that KIF has primitives that can
be used to define the CG context enclosures and that the metalanguage
facilities of KIF and CGs can be used to define McCarthy's lifting rules
as operations on those enclosures. Although we have not yet worked out
the formal details, Mike and I believe that we can reconcile the KIF and
CG mechanisms and use them to support the context facilities that McCarthy
and Guha want. There is still work to be done, but it seems to be a
nitty-gritty formalization rather than a major research breakthrough.
In previous notes, I had mentioned the technical report on conceptual
schemas that is being written by the ANSI X3H4.6 Task Group on
Conceptual Schemas for Information Resource Dictionary Systems (IRDS).
A 376-page draft has been finished, and it has been approved by the full
X3H4 Committee. It is now going through the final editing stage and
should be published as an ANSI TR in early 1993. I'll send a summary
of some of the major points in a future note.
In related developments, X3H4 has been interacting with the Knowledge
Sharing Effort, with ISO, and with other ANSI groups including X3T2 on
Communications and the newly formed X3H7 on Object Information Management:
1. During the past year, several people working on KSE have visited
X3H4 meetings to give presentations and discuss ways of cooperating.
Among them were Richard Fikes, Mike Genesereth, Bob Neches, and
Bill Swartout.
2. The TR discusses the importance of a logic-based approach and
recommends conceptual graphs as the basis for an "initial normative
language" for the IRDS conceptual schema. The Knowledge Interchange
Format (KIF) and the PDES SUMM (Semantic Unification Meta-Model) are
also considered as candidates for normative languages. As a result
of discussions with Fikes, Genesereth, and others, we believe that
it should be possible to develop a common semantic core for CGs,
KIF, SUMM, and related languages. There is considerable agreement
on the broad outlines of the common core, and working out the details
is a high-priority task.
3. In order to ensure that there is one general framework for
conceptual schemas instead of multiple incompatible frameworks,
the X3H4 Committee is collaborating with other ANSI Committees,
including X3T2 on Communications, X3H7 on Object Information
Management, and X3H6 on CASE Tools. In March 1993, X3H4 and X3H7
will hold a joint meeting in San Antonio to plan a common direction
for supporting and using object-oriented facilities in the IRDS.
4. Meanwhile, there is also considerable interest in conceptual schemas
at the international level. In March 1992, ISO held a special meeting
on conceptual schemas and data modeling facilities in Renesse, the
Netherlands. The six U.S. delegates strongly endorsed a logic-based
approach along the lines recommended by the X3H4 TR. Many of the
delegates from other countries were highly sympathetic to logic, but
there was also a vociferous minority that insisted on an SQL-based
approach to data modeling.
5. The next meeting of the ISO Special Working Group will be held in
Namur, Belgium, in December 1992. The U.S. delegates will represent
the X3H4 Committee on IRDS, the X3T2 Committee on Communications,
and the X3H7 Committee on Object Information Management. I plan to
go as a representative of X3H4, and Bob Neches is planning to go as
a representative of the Knowledge Sharing Effort.
The standards processes take years, and it is important to get input
>From the Knowledge Sharing Effort into the process right from the
beginning. The standards efforts already have the attention of a great
many people from the database, communications, and object-oriented
communities. This is a good opportunity for the AI community to get
directly involved in the mainstream of computer applications and help
to guide the future directions.
John Sowa