Browse Books

Reviewer: Susan Bridges

Sowa's latest book provides a unique view of the field of knowledge representation. This view is strongly grounded in philosophy and logic, with conceptual graph theory used as a unifying notation. Sowa describes knowledge representation as the application of logic and ontology to the task of constructing computable models for some domain. This book continues the tradition established in Sowa's first book, Conceptual structures [1], of integrating ideas from an amazing array of disciplines in a historically based, coherent, detailed, and disciplined manner. The first two chapters cover logic and ontology. The chapter on logic begins with a discussion of the development of logic by Greek philosophers, continues by covering a wide array of logic notations, and concludes with a section on “Unity Amidst Diversity.” This approach of discussing a need, reviewing the historical background, and presenting a unified view of the topic continues through the remainder of the book. For example, in the chapter on ontology, Sowa explains the need for ontologies in computational fields, discusses several of the major categorizations developed by philosophers, and then presents his top-level ontology and shows how it accommodates the views of others. The third chapter presents the major types of knowledge representations; the fourth chapter covers representing and reasoning about processes; and the fifth chapter discusses knowledge representation in complex environments that must accommodate purposes, contexts, and agents. Chapter 6, “Knowledge Soup,” covers a wide range of important knowledge representation issues that do not fit into the other chapters. The last chapter covers the important and often overlooked area of knowledge acquisition and sharing. Three appendices present notation and examples that are of critical importance in understanding the text. One of the great strengths of this book is the illustration of how different types of knowledge can be represented using a variety of techniques and languages. Many of the examples are presented in several notations or representations, including FOL, conceptual graph notation (both linear and graphical), CLIPS, SQL, and KIF. Many of them are useful for illustrating the strengths and weaknesses of different formalisms and notations. This book is of great value to practitioners of knowledge representation, to people from a variety of backgrounds who want to learn more about knowledge representation, and as a textbook in an upper-level course in knowledge representation. Although the book is relatively short by modern standards, it covers an immense amount of material in great detail. An instructor using the book for a course is faced with many decisions about which material to omit, and must constantly work to avoid becoming bogged down in too much detail. On the other hand, there are few alternative texts that provide the synthesis afforded by this book. Many exercises of widely varying difficulty are provided at the end of each chapter. Most of the exercises are thought-provoking, interesting, and informative. Unfortunately, the book has quite a few minor typesetting errors, which are quite distracting when it is used as a text. The bibliography is an extremely useful resource in and of itself. Sowa references and discusses much work of historical importance that is not mentioned by other authors and that can provide gateways to large bodies of relevant work that are otherwise easy to overlook. The greatest strengths of this book are its careful synthesis of the ideas from many disciplines and the author's wonderful writing style. The book was much anticipated and will become as widely read and cited as Sowa's first book.