Browse Prior Art Database

RESEARCH IN NATURAL LANGUAGE UNDERSTANDING (Report No. 1)

IP.com Disclosure Number: IPCOM000128805D
Original Publication Date: 1978-Dec-31
Included in the Prior Art Database: 2005-Sep-19

Publishing Venue

Software Patent Institute

Related People

W.A. Woods: AUTHOR [+3]

Abstract

The ARPA natural language understanding project at BBN is an effort to discover and develop techniques for dealing with large bodies of information in a command and control decision-making situation. Natural language understanding in the context of a " " knowledge-based intelligent display system constitutes a testbed for the research. Although this is an important application in its own right, the problems addressed in the research are general ones that cut across many other areas of application of rule-based and knowledge-based systems. The research project is primarily a design """ study to find and formulate fundamental techniques for efficiently solving a set of problems. The specific problems that we are investigating are those that we consider the most critical for a fluent communication interface to an information display system for command and control .~., decision-making. Although we approach this problem from the point " of view of natural language understanding, most of the techniques required would still be necessary for fluent and natural communication even for an artificial language designed specifically for the purpose. The characteristics that are most important are the ability for the user to omit detail that can be inferred by the system, and the ability to express the elements of the user's requests in the order in which they occur to him, without extensive forethought and problem-solving by the user in developing the specification of what he wants done. Also important is the <<...>> BBN Report No. 3742 Bolt Beranek and Newman Inc. capability of the system to prompt the user for detail that was omitted in his request when it is necessary to complete the specification of the request and is not inferable by the system. reference, dependent inferential To the largest extent possible, the user should be freed from the burden of remembering awkward and arbitrary syntactic conventions, from such anticipatory obligations as having to initially name any object that he may subsequently want to refer to, and from having to specify more detail in his requests than is necessary and natural. These objectives lead naturally to requirements for the system to handle the equivalent of anaphoric ellipsis, default assumptions, vagueness, context interpretations, and a fair degree of intelligent processing in determining the intent of a user's request. This latter, in turn, results in a need to handle efficiently large data bases of general world knowledge to support such inferences. Moreover, the nature of the information processing required for such inferences, together with the current revolution in the cost of processing elements in integrated circuits, lead to the desirability and likely necessity of parallel processing algorithms and architectures for handling such processing in real time. Consequently, the current research project involves an integrated attack on a number of related problems, each of which is a significant research problem in its own right. These include syntax and semantics of natural English, the interaction of BBN Report No. 3742 Bolt Beranek and Newman Inc.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 4% of the total text.

Page 1 of 34

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

RESEARCH IN NATURAL LANGUAGE UNDERSTANDING

W.A. Woods

Quarterly Technical Progress Report No. 1

1 September 1977 to 30 November 1977

ARPA Order No. 3414 Program Code No. 8D30 Name of Contractor: Bolt, Beranek & Newman, Inc. Effective date of contract: 1 September 1977 Amount of Contract $301,377 Contract No. N00014-77-C-0378 Contract Expiration Date: 31 August 1978 Short Title of Work: Natural Language Understanding Principal Investor: Dr. William A. Woods (617)491-1850 x361 Scientific Officer: Gordon D. Goldstein

Sponsored by Advanced Research Projects Agency ARPA Order No. 3414

This research was supported by the Advanced Research Projects Agency of the Department of Defense and was monitored by ONR under contract no. N00014-77-C-00378

The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the Advanced Research Projects Agency or the U.S GovernmentBBN Report No. 3742 Introduction . Bolt Beranek and Newman Inc.

TABLE OF CONTENTS

1. Knowledge-Based Natural Language Understanding . . . . . . . 1.1 The Role of a Knowledge
Network for an Intelligent Machine . . . . . . . . .

1.2 Parsing Situations . . . . . .

1.3 The Process of Situation Recognition

1.3.1 Factored Knowledge Structures .

1.3.2 Markable Classification Structures . . . . The Structure of Concepts . . . . . . . . . . . The
Need for Inheritance Structures . . . . .

1.4

1.5

1.6 The Taxonomic Lattice

1.7 An Example . . . .

1.8 Conclusions . . .

Bolt, Beranek & Newman, Inc. Page 1 Dec 31, 1978

Page 2 of 34

RESEARCH IN NATURAL LANGUAGE UNDERSTANDING (Report No. 1)

4 2. Structured Inheritance Networks 2.1 A Note on Notation . . . . . 2.2 Node Types . . . . . . .
2.2.1 Concept and Concept-node Types 2.2.2 Role- and SD-nodes . . . . . . . 2.3 Eplink Types . .
. . . . . 2.3.1 Intra-concept-links . . 2.3.2 Intra-role-links . . . 2.3.3 Intra-sd-links . . . . 2.3.4
Abstraction Hierarchies 2.3.5 Inter-role-links . . . 2.3.6 Inter-sd-links . . . . 2.4 SI-Net-1 Primitive Concepts . . . . . . . . . . 2.5 Meta-description and Procedural Attachment 2.6 The Individuation
Process . . . . . . . 2.7 Further Refinements . . . . . . . . . . . . . . . .. and Inter-concept-links

3. References . .

4 8 9 12 18 21 23 26 29 34 79

Introduction

The ARPA natural language understanding project at BBN is an effort to discover and develop techniques for dealing with large bodies of information in a command and control decision- making situation. Natural language understanding in the context of a

" " knowledge-based intelligent display system constitutes a testbed for the research. Although this is an important application in its own right, the problems addressed in the research are general ones that cut across many other areas of application of rule-based and knowledge- based systems. The research project is prima...