The Engineering of Knowledge-Based Systems: Theory and Practice/Book and 2 Disks - Hardcover

Gonzalez, Avelino J.; Dankel, Douglas D.

Hardcover

ISBN 10: 0132769409 ISBN 13: 9780132769402

Publisher: Prentice Hall, 1993

Synopsis

Emphasizing a hands-on approach for building knowledge-based systems, this guide covers both the theory and the practical applications of knowledge-based systems. KEY TOPICS: It presents the theoretical foundations of knowledge-based systems as a branch of artificial intelligence, and then concentrates on practical aspects in the development of a knowledge based system. For professionals working in the fields of knowledge-based engineering and artificial intelligence.

About this title

From the Publisher

This volume provides comprehensive single-volume coverage of both the theory and the applications of knowledge-based systems.

Excerpt. © Reprinted by permission. All rights reserved.

Preface

INTRODUCTION

In the early 1970s, researchers in Artificial Intelligence (AI) recognized that the general problem-solving methods and searching techniques developed over the previous ten years were insufficient to solve the difficult research and application-oriented problems of the day. They realized that what was required was specific knowledge about the particular, limited application domains of interest rather than broad general knowledge which applied across many domains. This recognition led to the development of knowledge-based (i.e., expert) systems. Since its inception, knowledge-based systems technology has grown into a dominant topic in the AI technical literature. From the simple concept in the mind of a researcher in a laboratory has emerged a rapidly evolving multimillion dollar industry of practical applications.

The arrival of this technology has had a major impact in industry and government. Knowledge-based systems are currently found in a broad diversity of fields ranging from accounting and banking to engineering design and medical diagnosis, with numerous applications currently under development. Applications range from simple knowledge-based systems like AUDITOR, an assistant to a professional auditor which evaluates a client's potential for defaulting on a loan Dungan, 1983; to I&W, which assists an intelligence analyst in predicting when and where an armed conflict will next occur Kiremidjian, 1983; to XCON, a VAX computer configuration system developed and used by Digital Equipment Corporation (DEC) McDermott, 1982. Systems are implemented on personal computers, dedicated workstations, and mini- and mainframe computers.

To support the diversity of applications and the range of their sizes, many support organizations have formed. These organizations include hardware manufacturers who build computers on which to develop systems efficiently, software developers who build tools to facilitate their construction, and consultants who assist not only in the selection of hardware and software, but also in the development of these systems themselves.

But why are knowledge-based systems so popular at this time? What is it about them that has grabbed the attention of business as well as the scientific community? The answer these questions requires an understanding of their structure as well as their relationship with other technological innovations of the present time. In this book we provide an in-depth examination of what knowledge-based systems are, how information is represented within these systems, and how these systems are constructed. Before we plunge into a description of knowledge-based systems, however, it would be good to see what has been accomplished in the field over the last 10 to 15 years. The amount of space in this preface dedicated to such a historical perspective does not do justice to the accomplishments which have occurred. Nevertheless, we shall have to be content with a quick review of a number of significant systems and a more in-depth glance at one major system now in commercial operation.

Numerous knowledge-based systems have been developed over the last two decades and many more are currently under development. Some of the "classic" systems merely served to show the applicability of this new technology in some new domain, while others evolved into commercially applied systems. The following descriptions illustrate the diversity of developed systems.

PROSPECTOR Duda, 1978 Assists geologists in identifying geological formations which may contain mineral deposits. It was developed by SRI International between 1974 and 1983 and did not mature into a commercial system.

XCON McDermott, 1982 Assists in the configuration of newly ordered VAX computer systems. XCON was developed by Digital Equipment Corporation (DEC) in conjunction with Carnegie-Mellon University. It is presently used internally by DEC.

MYCIN Shortliffe, 1976; Buchanan, 1984b Medical diagnostic system which determines the infectious agent in a patient's blood and specifies a treatment for this infection.

GUIDON Clancey, 1979, 1983 Instructional program teaching students therapies for patients with bacterial infections. GUIDON is a descendant of MYCIN and was developed as a research tool at Stanford University. It is not commercially available.

LES Scarl, 1987 Monitors and diagnoses the process of loading liquid oxygen (commonly referred to as LOX) into the space shuttle's main tank. It was developed by MITRE Corporation and NASA KSC (NASA Kennedy Space Center). LES never developed into a commercial application; however, the concepts developed in LES have been featured in another NASA developed knowledge-based system called KATE (Knowledge-based Autonomous Test Engineer). KATE is presently being applied to the environmental control system in the Orbiter Maintenance and Refurbishment Facility at the Kennedy Space Center.

ISIS Fox,1984 Generates a job shop schedule for a factory floor. It uses a technique called constraint-directed reasoning, which allows certain constraints to determine the sequence of operations needed to complete a job within the specified schedules. It evaluates the alternative sequences which are suggested to pick the best. It was developed by Carnegie-Mellon University and Westinghouse Electric Corporation in Pittsburgh and is presently being used by Westinghouse.

DELTA/CATS Bonissone, 1983) This is a rule-based diagnostic system for troubleshooting electric diesel locomotives. It was developed by the General Electric Company of Schenectady, NY, and is presently being used internally by the General Electric Company.

STEAMER Hollan, 1984 A simulation-based system developed to instruct naval propulsion engineering students in the operation of a shipboard steam propulsion plant. It was developed by the U.S. Navy in collaboration with BBN Corporation and is currently used in the training of naval personnel.

INTERNIST/CADUCEUS Miller, 1982 One of the. largest medical systems developed, INTERNIST assists the physician in making multiple and complex diagnoses in internal medicine. It was developed at the University of Pittsburgh and is not currently in commercial use.

COOKER AInteractions, 1985 Assists in the maintenance of soup-making equipment. Developed by Texas Instruments for the Campbell Soup Company, it uses a personal computer as the delivery platform and is currently in use within the Campbell Soup Company.

AUTHORIZER'S ASSISTANT Leonard-Barton, 1988) Assists the credit authorization staff determine the credit level for credit card customers. The system takes information from a number of databases and approves or disapproves a telephone request from a merchant to authorize a large purchase from a cardholder. It was developed by Inference Corporation and American Express Company and is currently used by American Express Company.

GENAID Gonzalez, 1986) Remotely monitors and diagnoses the status of large electrical generators in real time. It issues a diagnosis with a confidence factor whenever the machine is operating outside its normal operating conditions. It was developed by Westinghouse Electric Corporation with assistance from Texas Utilities Generating Company and Carnegie-Mellon University and is presently in commercial operation at various sites throughout the United States.

But how does the field of knowledge-based systems relate with AI? This question merits a closer look.

RELATIONSHIP TO AI

Very often, knowledge-based systems and AI are mistakenly assumed to be one and the same. One reason for the mistaken assumption is that the field of knowledge-based systems is the branch of AI which has, by far, seen the most success in terms of practical applications. Such thinking is fairly common among nontechnical people. The news media contributes to this phenomenon since the term artificial intelligence is more dramatic and sensational than knowledge-based systems. The former has often been used where the latter was intended.

While the use of knowledge has always been one of a set of techniques employed by AI practitioners, the knowledge used by the early Al systems tended to be very general in nature. The development of knowledge-based systems was marked by the realization that general knowledge was not sufficient to solve the difficult problems, and that high quality, domain-specific knowledge was the alternative.

But what exactly is meant by AI?

A good definition of AI is quite elusive simply because human intelligence is not completely understood. The various textbooks on AI provide diverse definitions emphasizing the different perspectives which their author(s) feel the field entails. Nevertheless, one general definition Tanimoto, 1987 which provides a good broad view of the field is:

Artificial intelligence is a field of study that encompasses computational techniques for performing tasks that apparently require intelligence when performed by humans. Such problems include diagnosing problems in automobiles, computers and people, designing new computers, writing stories and symphonies, finding mathematical theorems, assembling and inspecting products in factories, and negotiating international treaties. It is a technology of information processing concerned with processes of reasoning, learning, and perception.

Historically computers have excelled at performing rather simple, repetitive tasks such as complex arithmetic calculations or database storage and retrieval. What these repetitive tasks have in common is that they are algorithmic in nature. That is, they involve a precise and logically designed set of instructions which yield a single correct answer. Humans, on the other hand, excel at solving problems using symbols rather than numbers, such as when planning a schedule of tasks or understanding a poem. AI, in more specific terms, is the science which provides computers with the ability to represent...

Edition details

• Publisher: Prentice Hall
• Publication date: 1993
• ISBN 10: 0132769409
• ISBN 13: 9780132769402
• Binding: Hardcover
• Language: English
• Number of pages: 560