Browse Books

Reviewer: Alan F. Smeaton

Information retrieval (IR) is the field of matching users' vague and imprecise information needs against a corpus of unstructured information, retrieving parts of this collection that may be relevant to the information need, and presenting this information to the user. For many decades, this unstructured information had typically been collections of text documents; more recently, we have seen information retrieval applied to audio, image, and video information. As a discipline, IR has been evolving slowly and steadily for at least three decades, with the initial, small-scale applications in patent offices and online searching now being overtaken by online searching of all kinds of information by millions of corporate users and by the most public of IR applications, searching the World Wide Web. Formerly a backwater domain lurching between computer science and information science, information retrieval's star is on the rise because computing has evolved to make content-based manipulation of information increasingly crucial. It is in the context of this upsurge of interest that this book appears. Korfhage is a respected contributor to IR and is well qualified to write this book. It is written from an information science rather than a computer science perspective; this observation is meant as a compliment. The intended audience is not discussed in the book, but the companion Web site points out that the book is designed for an introductory graduate course on the concepts and methods of IR. It contains a good index and a useful glossary, and there are exercises at the end of each chapter. The companion Web site is a good idea, but it does not yet contain the promised hints and answers to exercises. The book covers the standard IR fare, including Boolean IR, the matching process, text analysis and indexing, measurement of retrieval effectiveness, relevance feedback, alternative retrieval techniques including IR and natural language processing and hypertext, interfaces to IR systems, and an emphasis on user profiles and their use. However, the coverage of retrieval from images or audio ( which are of increasing importance in IR) is weak or nonexistent. There is little or no coverage of some other important IR topics, such as clustering, NLP, TREC, and latent semantic indexing. There is little emphasis on implementation techniques, signature files, or partitioned searching, which is fine given the level to which the book is pitched, but then it is curious for Appendix A to be devoted to string searching algorithms, Appendix B to file structures, and 12 pages of the text to data compression. The content is deep and narrow, focusing on the basics of IR, rather than taking a scattergun approach of covering more topics. This may be fine for an introduction followed by an advanced IR course, but the number of topics covered could make this difficult. There are a few classic texts on information retrieval, and the present book is an update on some of these. It is less mathematical than van Rijsbergen [1] and more up to date than Salton [2]. Compared to more contemporary textbooks, it is less implementation-oriented than Frakes and Baeza-Yates [3] or Witten et al. [4] and better written, though less broad, than Kowalski [5]. The recent Readings in information retrieval edited by Sparck Jones and Willett [6] is also a useful graduate text. Despite all of these texts in the field, there is a place for Korfhage's book either as a graduate text or on the shelves of researchers who want to learn about information retrieval.