KIF SD-3 (Michael R. Genesereth)
Date: Mon, 2 May 1994 10:50:33 -0700
Message-id: <199405021750.AA13788@Sunburn.Stanford.EDU>
To: interlingua@ISI.EDU, kr-advisory@ISI.EDU, srkb@ISI.EDU
From: (Michael R. Genesereth)
Subject: KIF SD-3

As John Sowa mentioned, X3T2 recently voted its support of standardization
project proposals for CGs and for KIF.  John distributed the SD-3 (the
official proposal) for cgs.  I am enclosing the SD-3 for KIF below.


Proposal for a Project to Develop a Standard for
Knowledge Interchange Format (KIF)

1. Identification of the Proposed Project

1.1 Title

Knowledge Interchange Format

1.2 Proposer

Michael R. Genesereth
Computer Science Department
Stanford University
Stanford CA 94301
phone: 415-723-0324

1.3 Date Submitted

22 April 1994

1.4 Project Type - D

The development of the standard will be done within an X3 committee.  The
standard may also be progressed through ISO.  Upon establishment of an
international work item, the proposers will apply for a change of status to
international liaison project.

2. Justification of Proposed Standard

2.1 Needs

In this document, the word ``knowledge'' refers to information expressed in the
form of logical expressions, e.g. constraints, negations, disjunctions, rules,
quantified formulas, and so forth.  Logical form is useful for expressing the
definitions of concepts, physical and artifical laws, rules of thumb, and so

The number of programs capable of manipulating knowledge is growing rapidly,
and the rate of growth in this number is likely to increase over the coming
years.  Current applications include expert systems, natural language systems,
software specification and verification systems (in which specifications are
expressed as logical expressions), and computer-aided design systems (which
exchange partial design information and component behaviors in this way).  In
addition, many approaches to the interoperation of heterogeneous software rely
on the interchange of knowledge about the capabilities of programs and their
data (e.g. definitions for conceptual schemata, database view definitions, and
so forth).

While there are existing standards for information in many application areas
(e.g. SMTP for electronic mail, GIF for graphics, PostScript for documents),
there is no standard interchange format with the degree of expressiveness
needed by modern knowledge manipulation programs.  As a result, most such
systems today rely on ad hoc encodings of knowledge and, consequently, are
unable to interoperate with other knowledge manipulation programs.  There is a
growing need for a knowledge interchange format to remedy this situation.

2.1 Recommended Scope of Standard

The standard would be restricted to Knowledge Interchange Format (or KIF for
short).  The purpose of this format is the interchange of knowledge among
coresident or distributed computer programs.

KIF is NOT intended as a primary language for interaction with human users
(though it can be used for this purpose).  Different programs can interact
with their users in whatever forms are most appropriate to their applications
(for example frames, graphs, charts, tables, diagrams, natural language, and so

KIF is also NOT intended to be an internal representation for knowledge
WITHIN computer programs or within closely related sets of programs
(though it can be used for this purpose as well).  Typically, when a program
reads a  knowledge base in KIF, it converts the data into its own internal form
(specialized pointer structures, arrays, etc.).  All computation is done
using these internal forms.  When the program needs to communicate with
another program, it maps its internal data structures into KIF.

2.2 Existing Practice in Area of Proposed Standard

There are no national or international standards in this area.  In existing
practice, programmers communicate with each other to decide on knowledge formats
in an ad hoc fashion.

2.3 Expected Stability of Standard with Respect to Current and
Potential Technological Advance

The proposed standard would be based on the KIF specification (X3T2/93-041)
produced by the Interlingua Committee of the Arpa Knowledge Sharing Effort.  We
believe that the details provided in this document would lead to a very stable

First of all, the syntax of KIF corresponds to common practice in the Artificial
Intelligence community, the principal developers of knowledge technology.  The
semantics of KIF is based on results published in the archival literature of
Mathematics and Computer Science, and there are multiple implementations of
inference procedures for formats with this semantics.

Second, the specification of KIF has been widely distributed; papers on KIF have
been presented at various scientific conferences; and the format and its usage
have been the subject of extensive discussion in an open electronic forum.  In
addition, a document (X3T2/93-048 with attachments) announcing intent to
standardize and soliciting comments was circulated to 13 X3 committees; and
universally favorable comments were received.

3. Description of Proposed Project

3.1 Type of Document


3.2 Definitions of Concepts and Special Terms

Knowledge -- As used here, the term refers to information encoded in the form
of logical expressions and sentences, i.e. negations, disjunctions, rules,
quantified formulas, and so forth.  Simple data can usually be expressed in
the form of simple sentences (e.g. Joe is the father of John); knowledge
often requires more complex sentences (e.g. Joe is not the father of John, the
president of the company is the father of either John or Jill,  if x is the
father of y, then x is a parent of y).

3.3 Expected Relationship with Approved X3 Reference Models

KIF will enable enhanced efficiency and interoperation of products and services
that conform to standards derived from X3 and ISO reference models.

ANSI/SPARC 3-Schema Architecture.  KIF will support the specification and
interchange of conceptual schemas as defined in the ANSI/SPARC architecture.  In
this way, it will complete much of the program of work defined in X3 project

Reference Model for Open System Interconnection (OSI).  KIF will serve as an
efficient declarative format for specifying Application Service Objects (ASO's)
and application contexts at the ISO application layer.

Reference Model for Open Distributed Processing (ODP).  KIF will be a useful
format for interchanging the conceptual objects specified in the ODP Enterprise
and Information viewpoints.

3.4 Recommended Program of Work

Our primary jobs will be (1) to convert the KIF specification into the proper
format for a standard, (2) to provide a BNF account of the syntax of the format,
and (3) to outline operational considerations in the  use of the format (e.g.
the design of inference procedures).

3.5 Resources - Individuals and Organizations Competent in Subject Matter

The members of the Interlingua Committee of the Arpa Knowledge Sharing Effort
are experts in the area of knowledge technology.   Many of these individuals
have pledged to contribute their time to the establishment of a KIF standard. 
All will be available for consultation; some will join the X3 committee charged
with the development of the standard.

3.6 Recommended X3 Development Technical Committee


3.7 Anticipated Frequency and Duration of Meetings

Much of the discussion about this format to date has been carried out via
electronic mail, and this will continue into the future.  The responsible
committee will serve as the forum and approval body for KIF.  Approximately
four meetings per year will be required, each lasting two or three days.

3. Target Date for dpANS to X3

6 June 1995

3.9 Estimated Useful Life of Standard

The number of knowledge manipulation programs is increasing rapidly, and the
rate of growth in this number is likely to increase over the coming
years.  Consequently, there is a strong need for a standard in this area, and
this need will continue for the forseeable future.

It is expected that the standard will require periodic, upward compatible
extensions to keep pace with continuing developments in knowledge technology. 
The standard will be structured to enable timely amendments as the market for
KIF-compatible products develops and expands.

4. Implementation Impacts

4.1 Impact on Existing User Practices and Investments

The introduction of a standard for knowledge interchange will lead to software
products superior to those available today.  Users will see enhanced
interoperability (especially among heterogeneous systems) and greater
efficiency (due to the ability of systems to exchange partial information in a
timely manner and thereby avoid unnecessary work).
It will also be possible to convert ``legacy'' software to use KIF.  Vendors
will be able to ``wrap'' existing software or provide software ``agents'' that
act as transducers between the system's native format and KIF.  Technology
exists to support both of these approaches.

4.2 Impact on Supplier Products and Support

Over the coming years, there will be enormous growth in the number of knowledge
manipulation systems.  Completion of this project will provide the vendors of
these systems with the necessary standards to guarantee interoperation.

The standard will also open the marketplace for ``knowledge products'',  i.e.
bases of encoded knowledge analogous to today's databases.

Along with the market for knowledge will come the opportunity to market support
tools of various sorts, e.g. inference programs, personal ``agents'', graphical
and natural language user interface programs for KIF, translators between KIF
and native formats, and translators between different KIF ``ontologies''.

4.3 Techniques and Costs for Compliance Verification

Compliance with the syntax of KIF is readily and automatically determinable, as
its grammar is simple and clearly defined.

Semantic compliance is equivalent to logical ``soundness''; an inference program
complies with the semantics of KIF so long as it does not derive any conclusions
(sentences written in KIF) that are not logically implied from the available
premises (sentences written in KIF).  The concept of logical implication is
fully defined in the KIF specification, and compliance is subject to
in the same way that a programming language interpreter or compiler is
subject to

4.4 Legal Considerations

There are no known copyright, patent, or restraint of trade issues associated
with this project.  The KIF specification itself is in the public domain.  

Public interest will be served by assuring that meetings remain open to the
public, which has a direct interest in the standard.  All documents paroduced
by or submitted for consideration by the committee responsible for the proposed
standard will be available for public review through the X3 Secretariat.

5. Closely Related Standards Activities

5.1 Existing Standards

All existing data exchange standards are specific to particular domains, e.g.
graphics, electronic mail, and so forth, or particular tasks (e.g. software
specification).  None have the expressiveness of KIF, e.g. the ability to
express negations, disjunctions, rules, quantified formulas, and metalevel

5.2 X3 Standards Development Projects


5.3 X3/SPARC Study Groups

Database Systems Study Group.  KIF can serve as a communication format for
advanced object-based distributed processing agents; and, in this way, it may
support the database industry trends currently under review by this study group.

5.4 Other Related Domestic Standards Efforts

1015-L -- ``Z Specification Language (Z)''. The TAG for this group was recently
assigned to X3J21.

1043-L -- ``Conceptual Schema Modelling Facilities'', the liaison project
to the generic conceptual schema modelling
facility (CSMF) standards project under ISO/IEC JTC1/SC21/WG3.  The TAG for
this group was recently assigned to X3T2, and KIF input to the international
standards community may come through this vehicle.

Project 689-D ``Conceptual Schema Specification for Data Interchange''

Project 988-D `Information Resource Dictionary System (IRDS) Normative Schema''

Proposed new project to establish a standard for Conceptual Graphs.  See
X3T2/94-043.  Conceptual Graphs provide a graphical representation for knowledge
of primary use in human interaction.  By contrast, KIF is a format for the
interchange of knowledge among computer systems.  These two formats are
complementary, and so the proposals are non-competitive and non-redundant.

5.5 ISO Standards Development Projects


5.6 Other Related International Standards Development Projects

1015-L -- ``Z Specification Language (Z)''.

1043-L -- ``Conceptual Schema Modelling Facilities''

5.7 Recommendations for Coordinating Liaison


5.8 Recommendations for Close Liaison