Knowledge Systems Design John K. Debenham Prentice Hall, 1989 Reviewed by: Vibhu Mittal U S C / I n f o r m a t i o n Sciences Institute 4676 Admiralty Way Marina del Rey, C A 90292-6695 mittal@isi.edu The author introduces the book in the preface as one meant for people interested in building expert systems and deductive databases, and in particular for those interested in building maintainable expert systems. He states that the book is not meant for a beginner, but is "designed for the educated, knowledge processing specialist". The book is slanted towards working professionals, rather than academics, a stance which is indicated by his statement in the preface where he states that "[the book] should be of prime interest to those large corporations with a commitment to building a large, integrated, corporate knowledge base". The book is well written, with numerous figures and examples to illustrate the author's points. A logic programming formalism is used to express the examples in the text. The book is relatively self-contained in that it contains a chapter on the major concepts behind the use of computational logic as a programming and a database language. After laying the basic foundations for understanding computational logic, the author illustrates how the choice of various parameters in structuring the knowledge representation can affect the performance of the overall system. Using small examples, the book illustrates how the choice of relations used can infiuence the conclusions that can be derived, and how the same knowledge can be represented in different ways. Throughout the book, the author stresses the need to make explicit, as much as possible, all the relations and the dependencies between the data items used in the system. The book also describes the use of "normal forms" to try and ensure that all the information about a concept is represented at only one place in the system. The use of these two principles in constructing the knowledge base results in a system that is easy to understand and explain, and consequently easy to modify and maintain. To this end, the author devotes a considerable amount of effort to describe a framework for representing knowledge in "normal" forms. The author also categorizes information about an application into three classes: data - the fundamental, primitive concepts in the system, i n f o r m a t i o n - the implicit, functional associations between the data, and k n o w l e d g e - the explicit relations defined in the representation. Using these classes, he then proceeds to enumerate rules to re-structure knowledge represented in these three categories in "normal" forms, which are basically extensions of the classical normal forms in relational algebra. This seetion is very useful as it extends the concept of normal forms beyond those used in relational algebra, and lays down a set of ground rules, following which will lead to a representation which is easy to maintain and modify. The book also has sections on knowledge acquisition, analysis and knowledge base engineering, wherein the author uses the above approach to illustrate how a good representation can make these tasks simpler and easier. The book does suffer from some weaknesses though, the major one being that the author does not consistently address the same audience all the time. The section on logic programming would seem to indicate that the author does not expect his audience to be well versed in lone. The section on logic programming contains information that would normally be available in an undergraduate AI course. However, his treatment of the classical (database) normal forms would seem to indicate that the author is addressing a naive database person. Even though the author states in the preface that the book is not meant for a beginner, and that the user is assumed to have read most of the current review literature, he spends some time in explaining terms such as "expert systems", "knowledge base systems" and "knowledge acquisition" at very basic levels. The chapter on knowledge acquisition for example would be useful to somebody who had never been exposed to the term before. At spots like these, it would have been useful to have pointers to references where one could get more information, such as current literature in the field. However, the book offers no such references, or mention alternative approaches to particular problems, other than the author's OWn. The book also uses some terms like "B-R Diagrams" and the "Entity-Relationship Approach" without offering any description of what these might be. The author then proceeds to make use of these in some examples (from which the meaning of these terms may be gleaned). (It turns out that "B-R" stands for"Binary-Relationship"). These points, together with the fact that the examples used in the book are drawn from a business dataprocessing environment ("Salary records", "Personnel field", etc) would seem to indicate that the book is aimed mainly at the database processing specialist with an interest in AI, who would like to know more about knowledge representation. However, the book is well written, with many worked-out examples to clarify points and is easy to read (if a little slow in some parts). And it does offer a new insight to analyzing information representation using an extension of the principle of normal forms in classical databases. S I G A R T Bulletin Vol. 2, No. 1

/lp/association-for-computing-machinery/book-review-knowledge-systems-design-by-john-k-debenham-prentice-hall-eGHJpBprwc