Mining the Web: Discovering Knowledge from Hypertext Data

The field of relevance algorithms for the web is still relatively new and the author provides a clear, informative introduction to the still-developing field. Many references to real problems are discussed, and the author avoids needless use of equations or symbolic logic when a simple textual explanation is more appropriate. This is the book that the authors of "Modelling the Internet and the Web" should have written. Avoid that book, it is a confusing disaster.

Probably not a book you're going to put on your coffee table, but if you've got any interest in this subject matter at all this is a book worth having. You can flip it open to just about any page and find something interesting. Most of the descriptions in this book move from general to specific, so you can jump around from chapter to chapter getting an overview, or dig more deeply when you want more detail. The references at the end of each chapter are also very useful. Whether you want a survey of the field or are trying to implement something specific, this book is a valuable resource.

Great book give lots of tips and teaches some great things for online web use or development.

very interesting book.. data mining is one of the most interesting fields in computer science and this book covers very interesting parts I enjoyed reading it

Executive summary: This is a fabulous book, written with care andprecision, easy to read yet covering in detail a wide variety ofthe most beautiful and promising developments in data mining andmachine learning as it relates to the World Wide Web, including aprescient vision of where the field is headed in the future.More detail: There are science authors who are clear experts intheir field, yet have trouble communicating their knowledge. Thenthere are science authors who write with clarity, but achieve itby dumbing down technical details to cater to a broad readership.Finally, there are authors who are experts and leaders in theirfield, who are actively contributing to the forefront of research,who are excellent writers, and who can communicate complexconcepts to a diverse audience with acumen, without glossing overimportant details. Soumen Chakrabarti is one such author. "Miningthe Web" is a stunning achievement. It is an excellent summary ofthe past decade or so of research in the area, covering nearly allof the important bases, including the machinery of Web crawling,Web information retrieval (i.e., search engines), clustering,automated classification, semi-supervised approaches, socialnetwork analysis, and focused crawling. Though Chakrabarti himselfhas contributed prominently to the field, this book is not at allthe vehicle for self-promotion that other specialist textssometimes feel like. The book should be valuable to newcomers,students, and experts alike, and could certainly serve as anexcellent course textbook. High-level concepts can be grasped withlittle mathematical background, yet more technically sophisticatedreaders will not be disappointed: most topics do include rigorouscoverage. The text is well organized, well written, and wellconceived. It's design, including generous and illuminatingfigures and illustrations, possesses an artist's touch, perhapsnot surprising given that Chakrabarti designs his own fontlibraries in his (apparently scant) spare time. It's hard toimagine where Chakrabarti found the time to write such acomprehensive and thoughtful book, but I'm not asking anyquestions: I'm thrilled with the outcome. The book is a must-havereference for anyone working in -- or aspiring to work in -- thecrossroads of Web algorithmics, data mining, and machine learning. David M. Pennock Senior Research Scientist, Overture Services, Inc. [website]

This book is an excellent introduction to a number of techniques in information retrieval, machine learning, data mining, network analysis and the application of such techniques to the Web. It discusses many research issues as well as provides practical insights into constructing Web mining tools and systems. Chakrabarti has brought the wisdom of researchers in the area of Web mining to a wider audience. I think the book will prompt the development of new courses for graduate as well as senior undergraduate students.The first part of the book deals with interesting practical and theoretical issues related with designing large-scale Web crawlers and search engines. Chapter 4 and 5 are a good introduction to various unsupervised and supervised learning methods. Although proper understanding of advanced methods like the LSI are possible only through adequate foundation in linear algebra (you can get only a flavor of the technique in the book). Part III of the book is my personal favorite. It has detailed description of various social network analysis methods, some of which have been applied by modern search engines like Google. Focused crawling, an area that the author has personally shaped, is also explained well. The book ends with a brief peek into the future of Web mining.The comprehensive yet easy to read nature of the book makes it a valuable addition to my shelf. It is hard to find a comparable book in the area of Web mining.

I still gave it 5 stars though the effective page number is 326. There are mainly 3 sections in the book --- the first section is 79 pages walks you thru the basic structure of a web search engine, the 2nd one talks about the learning process (clustering, classification and so on), yes, I know it is AI related stuffs, but this book does not have too much equation and is quite readable. From page 203 is the 3rd section --- application which includes page ranking and other interesting topics.

This book is one of the best computer science textbooks i have ever seen. Apart from the wealth of information and discussion on specific WEB crawling and data mining (chapters 2, 3, 7, 8), chapters 4, 5 and 6 constitute a wonderful summary of machine learning in general.The book's discussion of unsupervised learning (the EM algorithm, advanced algorithms in which the number of clusters is not known in advance), supervised learning (Bayesian networks, entropian methods, SVMs), semisupervised learning, co-training and rule induction is extraordinary in that it is short, intuitive, does not sacrifice mathematical rigor, and accompanied by examples (all taken from information retreival over the web).