Browse Books

Reviewer: Dimitrios Katsaros

This provides a good source of information for a young and significant research area: mining the Web's content and structure. A prominent researcher in that community wrote the book. Three parts make up the book. The first part contains introductory material on Web crawlers, as well as fundamental concepts from the field of information retrieval. The second part, which makes up the bulk of the book, focuses on machine learning techniques for hypertext. Essentially, it contains methods appropriate for finding statistical relations between attributes extracted from Web documents. Finally, the third part contains a collection of applications that draw on the techniques discussed in earlier chapters. The main strength of the book is in the second part, and the first chapter of the third part. The three chapters of the second part, which are devoted to clustering and to supervised and semi-supervised learning, manage to provide a thorough presentation of the issues and solutions they cover. The explanations provided, and the language used, reveal the author's deep understanding of the field. I really enjoyed reading the first chapter of the third part, which addresses algorithms for mining the Web's link structure. The author achieves a unified treatment of the presented methods (hits, page rank, and so on), and provides clear and documented arguments for each method's shortcomings and benefits. As for negative aspects in this book, the omission of a Web usage mining area should be mentioned. Although this area has borrowed some techniques from the more traditional areas of the data mining field (for which excellent books already exist), it has made some significant contributions of its own to the general area of data mining, developed to deal with its own particularities. In terms of physical presentation, a partly unsuitable characteristic of the book is the paper-like appearance of some chapters in the third part. These contain many graphs and performance measurements, which are not customarily included in textbooks. Overall, I believe this book is a significant contribution to the literature on data mining, and is a must read for anyone who is involved, as a researcher or practitioner, in the practice of extracting interesting patterns from the Web. I undoubtedly recommend it. Online Computing Reviews Service