Intended as an introductory textbook at the undergraduate level, this book covers expert systems for students of computer science. Chapters address “Logic and Resolution,” “Production Rules and Inference,” “Frames and Inheritance,” “Reasoning with Uncertainty,” “Tools for Knowledge and Inference Inspection,” and finally expert system toolkits (OPS5, LOOPS, and CENTAUR).
Many books have been written on expert systems, but few have attempted to be in the classical textbook format. This book makes a serious attempt at that form. It tries to distinguish itself from other AI textbooks by focusing exclusively on methods and techniques that are known to be practical and powerful for expert systems development. The chapters are liberally sprinkled with thought-provoking examples, and exercises and suggested readings are included at the end of each chapter.
The core strength of the book lies in its three chapters on three well-known expert systems building paradigms: logic and resolution, production rules and inference, and frames and inheritance. Using the knowledge/control model of an expert system consistently through these chapters, the authors go into a fair amount of detail on each approach. These three methodologies are fairly stable, and the authors have done them justice by covering this material in a methodical fashion.
The material in the following two chapters, on reasoning with uncertainty and tools for knowledge and inference inspection, is current and will serve as useful starting material but not the final word. These two areas are continuing to absorb new AI techniques, and any course on expert systems will have to augment the material in these chapters with more recent results from journal and conference papers.
The final chapter is on OPS5, LOOPS and CENTAUR. The intent of the authors is probably to provide insight into three well-known systems for building expert systems. These three systems definitely deserve description, since they were among the earliest practical implementations of various knowledge/control models, and they are presented nicely. This chapter might have been better if it had also described the technical innards of the more popular expert systems toolkits that are commercially available and are being used, however.
Almost any textbook on a new and evolving subject like expert systems will need additional material to keep a course current. This book does a good job of encompassing the stable core material in a complete and consistent fashion. The examples and exercises make the reader think and realize the potential drawbacks as well as advantages of the techniques in question. Check this book out before you decide on a textbook for your next expert systems course.