Building Large Knowledge-Based Systems: Representation and Inference in the Cyc Project

✍ Scribed by Douglas B. Lenat, R.V. Guha

• Publisher: Addison-Wesley Publishing Company, Inc
• Year: 1989
• Tongue: English
• Leaves: 396
• Category: Library

✦ Synopsis

"The Cyc Project is outrageously ambitious; it is actually doing what AI has been theorizing about for three decades. In an entertaining fashion, Lenat and Guha describe Cyc's collision of software development and philosophy in loving engineering detail." - Patrick J. Hayes, Principal Scientist, Xerox Palo Alto Research Center; President-Elect AAAI

Knowledge-based systems today lack common sense. As a result, the coming decade may see horrifying catastrophes blamed on software. The Cy Project addresses that limitation. Is is the world's first attempt to encode the hundreds of millions of facts and heuristics that comprise human consensus reality. This book is a mid-term report on that 1984-1994 effort going on at MCC in Austin and Palo Alto. Through its various representation language techniques, inference schemes, and ontology of common-sense knowledge, Cyc will change the nature of AI research, and the scope and nature of what AI applications can do.

As a system engineer, programmer, or project leader, one needs to better understand the dangers of relying on AI programs that have only a thin veneer of competence. The discoveries and techniques of the Cyc Project provide answers for anyone faced with bringing knowledge-based systems beyond their fragile limits and into the 1990s.

Building Large Knowledge-Based Systems: Representation and Inference in the Cyc Project provides:

• the authoritative status report on Cyc - the size, scope, and organization of its knowledge base; and a look ahead
• a detailed understanding of the CycL representation language, from frames to constraints to meta-level control
• an explanation of how and why Cyc draws on two dozen different built-in inference mechanisms - the familiar, unfamiliar, and exotic - each with its own Truth Maintenance Systems
• solutions to encoding representation "thorns", such as time, space, substance, and causality - techniques that expand the boundaries of knowledge-based systems

Douglas B. Lenat is head of the Cyc Project and the Principal Scientist at the Microelectronics and Computer Technology Corporation (MCC). R.V. Guha, a mechanical engineer and computer scientist, is coleader of the Cyc effort at MCC.

✦ Table of Contents

Acknowledgements
Preface
1. The Cyc Philosophy
1.1. Thin Ice
1.2. Overcoming Brittleness
1.3. Falling Back on General Knowledge
1.4. Analogizing to Far-flung Knowledge
1.5. The Representation Trap
1.6. Ontological vs. Knowledge Engineering or Why Building a Huge KB is Different from Building n Small KBs
1.7. Building Cyc
2. Overview of Cyc
2.1. The KB Itself
2.2. The Interface
2.3. The CycL Representation Language - Introduction
3. The CycL Representation Language
3.1. The Rationale Behind the Design of CycL
3.2. Details of the Frame Representation and the CycL Constraint Language
3.3. Details of Inferencing in CycL
3.4. Getting, Putting, and Metalevel Inference
3.5. Neats versus Scruffies: A Digression
4. Representation: What and Why
4.1. Who Gets To Be Somebody, or, The Cosmic Cookie Cutter
4.2. Why Are Categories Useful?
4.3. AttributeValues versus Categories
5. A Glimpse of Cyc's Global Ontology
5.1. The General Categories in Cyc
5.2. Event, Process, EventType, and ProcessType
5.3. Slot - and Other Useful Sets of Slots
6. Representational Thorns, and Their Blunting
6.1. What the Universe Is Made of - Stuff and Things
6.2. Time
6.3. Causality and Related Issues
6.4. Intelligent Beings
6.5. Copying with Uncertainty
6.6. Structures (including Scripts)
6.7. Space
7. Mistakes Commonly Made When Knowledge Is Entered
7.1. Confusing an IndividualObject with a Collection
7.2. Defining a New Collection versus Definitng a New Attribute
7.3. Attaching Meaning to a Unit's Name or Its #%english Slot
7.4. Overspecific/Overgeneral Information
7.5. Accidental Duplication of Effort
8. Conclusions
8.1. Conclusions about the Cyc Philosophy/Motivation
8.2. Conclusions about the Global Ontology
8.3. Conclusions about Common Mistakes in Using the System
8.4. The Current State of Cyc
8.5. The Future
References

✦ Subjects

Artificial Intelligence, Expert Systems, System Design, CYC Project