Review of Knowledge Representation: Logical, Philosophical, and Computational Foundations
Guarino, Nicola, AI Magazine
Books Recently Received
Metal and Flesh by Ollivier Dyens
Similarity and Categorization by U. Hahn and M. Ramscar
Please visit www.cis.ohio-state. edu/~chandra/books-needing-- reviewers.htm for the assignment status of these books as well as for any new additions. If you are interested in writing a review, contact chandra@cis. ohio-state.edu.
Knowledge Representation-Logical, Philosophical, and Computational Foundations, John F. Sowa, Pacific Grove, California, Brooks/Cole, 2000, 534 pp., ISBN 0534949657.
Like his previous book, Conceptual Structures (Addison-Wesley, 1994), (Sowa's new book is a unique blend of philosophy, computer science, linguistics, and mathematics. It is intended to be a "general textbook of knowledge-base analysis and design" (p. xi), but I see it more as a collection of intellectual provocations, stimulating exercises, and cross-disciplinary comparisons for the appreciation of a more mature audience. Its great strength is recognizing the need for an interdisciplinary approach, and the attempt at presenting the logical and philosophical foundations of knowledge representation under a unified view. Its great weakness is a lack of consistent rigor, which is needed in a textbook for newcomers to a subject.
After some historical remarks and a first introductory chapter devoted to logic, Sowa immediately attacks the hard problems involved in choosing ontological categories, which lie at the heart of any knowledge representation project. This chapter is the densest (and most problematic), where the basic distinctions used and refined throughout the book are presented. Then the author overviews the main representational paradigms and delves into the muddy waters of times, events, processes, purposes, contexts, and agents. The subtle issues related to the limits of logic when dealing with vagueness, uncertainty, and ignorance are also addressed in some detail, in the context of what Sowa calls the knowledge soup, "a collection of signs-images, symbols, words, and concepts with associated feelings" (p. 394); this discussion ends with a synthesis of Peirce's and Saussure's accounts of semiotics, the study of signs. A final chapter overviews the problems and the techniques for knowledge sharing and acquisition, including a discussion about relationships between different representation systems. The book is supplemented by an extensive appendix, which includes a sample ontology (with a preliminary axiomatization of the top-level categories), an extended example, and answers to selected exercises.
My feeling is that this book is meant for adults only. I mean that although Sowa's first book was suitable for a large audience (and I myself still recommend it to knowledge engineering newcomers), this one is less systematic and more problematic, reflecting years of passionate inquiry into the deepest foundations of conceptual analysis and knowledge representation. What emerges is a vivid picture of the author's peculiar view of history of logic and philosophy, which is deeply intertwined with the analysis of the big problems of knowledge representation and knowledge engineering. Unfortunately, clarity and formal rigor are sacrificed in this highly interdisciplinary attempt, so that a good background in logic, philosophy, and knowledge engineering is necessary to avoid being puzzled by the many confusions, inaccuracies, and formal glitches and to extract the useful lessons.
This tolerant and open-minded attitude is especially necessary when reading the first few chapters of the book, the tough ones. In chapter 1, a simple sentence is chosen to introduce the reader to the problems of knowledge representation, "every trailer truck has 18 wheels" (p. 12). This example is analyzed as saying that if x is a trailer truck, then a particular set s of cardinality 18 exists, whose members are both wheels and parts of the truck. Now, this formalization doesn't exclude the case where the truck has more than 18 wheels, because if it had, say, 20 wheels, then the set with 18 wheels would still exist. …