Resolution on common principles
sowa@watson.ibm.com
Message-id: <9201210000.AA27995@quark.isi.edu>
Date: Mon, 20 Jan 92 18:43:11 EST
From: sowa@watson.ibm.com
To: JFULTON@atc.boeing.com
Cc: skperez@mcimail.com, roger@ci.deere.com, scott@ontek.com,
jhanna@mcimail.com, tony@ontek.com, msingh@bcr.cc.bellcore.com,
jonesm@atc.boeing.com, rls0028gks@atc.boeing.com, speyer@mcc.com,
tyler@cme.nist.gov, mzemanko@note2.nsf.gov, srkb@ISI.EDU,
kr-advisory@ISI.EDU
Subject: Resolution on common principles
Jim,
The resolution that was passed by ANSI X3H4.6 stated our assumptions
about the logic-based approaches to knowledge and information sharing.
The intent of the resolution was to announce to ISO, PDES, ANSI, and
other groups that all of us who are proposing logic-based approaches
have more in common than our differing notations might suggest.
At the ISO IRDS meeting in Tokyo last November, certain people got
the impression that the US TAG "couldn't agree among themselves."
If we allow that impression to go uncorrected, we could all lose.
I believe that we should make four points very clear:
1. The groups that are proposing logic as a basis for knowledge and
information sharing all agree that logic is the only solid basis
for such sharing.
2. The issues on which they disagree are negotiable points of detail.
3. The differing notations are all capable of expressing exactly the
same semantic core, which should be the essence of any standard that
may be adopted. Any or all of the differing notations should be
available as options.
4. No one should interpret a difference in notation as a disagreement
about fundamental semantic issues.
Following are my comments on your comments, and I hope after a couple
of iterations, all the people on these various committees can converge
on a common position.
>> These are the working assumptions for establishing a coalition for
>> logic-based standards, including ANSI X3H4, PDES SUMM, the DARPA-
>> sponsored knowledge sharing effort, and any other groups that wish
>> to join:
>>
>> 1. Symbolic logic, in its various forms, is the best developed, best
>> understood, and most precisely formalized knowledge representation
>> language available.
>>
> Comment: Although the Dictionary Methodology Committee agrees in
> principle with this assertion, our objectives do not at this time
> require the standardization of any concrete form of symbolic logic
> as a tool for modeling particular application domains.
When you say "concrete form", I am not sure what you mean. If you
mean "concrete syntax", I would agree that we should support any syntax
that is capable of expressing the agreed-upon semantics. But I think
that it is essential to establish a firm agreement on the semantic
base. Otherwise, we would have no basis for information interchange.
>> 2. Several different groups have adopted some version of symbolic
>> logic as a basis for defining the semantics of information.
>> These include the ANSI X3H4 IRDS committee with their Normative
>> Language, the PDES Dictionary Methodology Committee with their
>> Semantic Unification Meta-Model, and the DARPA-sponsored Knowledge
>> Interchange Format.
>>
> Comment: We are not certain that each of these groups is using
> symbolic logic for the purpose of "defining the semantics of
> information" in any common sense. KIF, for example, appears to be
> designed as a means of modeling specific domains. The SUMM, on
> the other hand, is explicitly intended to be used as a meta-model
> for the unification of models; a concrete implementation of SUMM
> would be a meta-language within which to describe other languages
> and their semantics. To this end the SUMM enriches symbolic logic
> with the extensions needed to accomplish that goal, including
> set theory and second-order objects. We have not yet seen a
> specification of the IRDS Normative Language.
All of these requirements are interrelated, and a version of KIF
that is rich enough to serve as a "means of modeling specific domains"
would have a very high overlap with a system that is intended as
"a metalanguage within which to describe other languages and their
semantics." For the IRDS Normative Language, we are using two
concrete syntaxes with a formally defined mapping between them:
sorted predicate calculus and conceptual graphs. Both of them
include the ability to represent sets and second-order objects.
>> 3. All of these groups have similar goals and requirements and have
>> developed versions of logic with very similar semantics. We do
>> not foresee any technical obstacles that would prevent them from
>> converging on a semantically equivalent foundation.
>>
> Comment: We would say rather that these groups may share the
> objective of being able to unify and integrate models of all types,
> although each group has other differing (but not incompatible)
> goals and objectives. We also agree that formal logic provides
> one of the primary foundations of that objective. However, these
> groups seem to have varying interpretations of how this can be
> achieved. We share the belief that there are no insuperable
> technical obstacles to this agenda.
I agree that all these groups have a high overlap in their objectives
and requirements. In order to promote convergence, I think that it
is important for all of us to state our objectives and requirements
explicitly so that we would have a basis for evaluating various proposals
for meeting those requirements.
>> 4. Therefore, we endorse the efforts of these and other groups to
>> work on logic-based standards for information interchange.
>> We further propose that such groups form a coalition for
>> logic-based standards that would all have a common semantic
>> foundation. Each group would remain free to develop any concrete
>> syntaxes that may be useful or convenient for their purposes.
>>
> Comment: We believe that such a coalition would be a very good
> step toward providing a framework that would enable the effective
> exchange of knowledge. We suggest that the work on this subject
> would be effectively organized on an international basis as a part
> of the recently approved ISO/IEC JTC1 SC21 Question on the SUMM
> and data models. (Note that the question does not assume that the
> SUMM in its current form is the answer. It asks what the problem
> is that the SUMM addresses and whether there should be an IRDS
> content standard to solve those problems. We agree with ANSI
> X3H4 that the ultimate answer will be some application of symbolic
> logic.) We suggest that the US perform its work on logic-based
> standards as part of the work organized by the US TAG on this Question.
For those of us who are working on logic-based proposals, we need a
very detailed list of requirements that can be used to evaluate one
feature or another. But to announce to the outside world that we
are fundamentally in agreement, we need a short, simple resolution
that says "We are all working towards a common logic-based approach."
The resolution passed by X3H4.6 was intended as a fairly short
statement of a consensus. If you or anyone else can think of a better
way to state the common principles, please do so. But I don't think
we should waste too much time on saying that we are basically in
agreement. Let's do that soon, and get on to the details that remain
to be negotiated.
John Sowa