Text Mining: Predictive Methods for Analyzing Unstructured Information (Hardcover)

by Sholom Weiss (Author), Nitin Indurkhya (Author), Tong Zhang (Author), Fred Damerau (Author)
Key Phrases: covering rule set, multiword features, spam classifier, System Deployment, Wall Street Journal, Ask Jeeves (more...)

Editorial Reviews

Product Description
The growth of the web can be seen as an expanding public digital library collection. Online digital information extends far beyond the web and its publicly available information. Huge amounts of information are private and are of interest to local communities, such as the records of customers of a business. This information is overwhelmingly text and has its record-keeping purpose, but an automated analysis might be desirable to find patterns in the stored records. Analogous to this data mining is text mining, which also finds patterns and trends in information samples but which does so with far less structured--though with greater immediate utility for users--ingredients. This book focuses on the concepts and methods needed to expand horizons beyond structured, numeric data to automated mining of text samples. It introduces the new world of text mining and examines proven methods for various critical text-mining tasks, such as automated document indexing and information retrieval and search. New research areas are explored, such as information extraction and document summarization, that rely on evolving text-mining techniques.

Product Details

• Hardcover: 236 pages
• Publisher: Springer; 1 edition (October 25, 2004)
• Language: English
• ISBN-10: 0387954333
• ISBN-13: 978-0387954332
• Product Dimensions: 9.3 x 6.1 x 0.7 inches
• Shipping Weight: 1.1 pounds (View shipping rates and policies)
• Average Customer Review: 4.0 out of 5 stars See all reviews (5 customer reviews)
• Amazon.com Sales Rank: #662,173 in Books (See Bestsellers in Books)

Customer Reviews

Most Helpful Customer Reviews

11 of 15 people found the following review helpful:

5.0 out of 5 stars An excellent introduction, July 7, 2005

By	Michael Stigall "JavaCat" (Mountain View, CA) - See all my reviews (REAL NAME)

I found the book informative and timely. The book describes the algorithms with psuedocode, and this made it possible for me to apply the algorithms to a legacy structure using another language within a few days of finishing the book.

The book's software (available from their website) requires XML formatted documents for input.

13 of 18 people found the following review helpful:

1.0 out of 5 stars it is 200 pages thick, April 5, 2007

By	Zhefu Zhang (USA) - See all my reviews (REAL NAME)

The authors (4 guys) tried to cover all IR big words as much as they can, and ended up with the 221 pages book. Let's take one example, inverted index takes 1.5 pages: It says inverted index is a table with the key-pair. The key is all the keywords scanned from the source, and the value is the document and word position (key in that document), period. IMHO, it is apparent facts that inverted index is like this way. But in practical world algorithm, it is much more complicated than a table, for example, how to incremental fill the index, how to sync between multiple backup copy, how to blabla. And even THAT google paper is more useful than it on this area. People may argue it is a comprehensive introduction book, well, then try Gerald's classic book. The whole impression is like I am reading a C++ programming book which spends 10 pages talking K&R's from Bell, how long they had been there, etc ...
I spent about 1 hr scanning the whole book without much left on my brain . Considering the price 69 bucks, I have to give it 0 on performance/price.

6 of 8 people found the following review helpful:

5.0 out of 5 stars an idiot savant, statistical viewpoint , October 26, 2006

By	W Boudville (Terra, Sol 3) - See all my reviews (TOP 50 REVIEWER)    (REAL NAME)

The authors give an excellent review of how matters stood in 2004, regarding text mining. The approach of the book is to minimise linguistic and semantic analysis. Instead, it looks more at the statistics of words (tokens) in documents. By using various such methods, they offer an automated way to classify documents. When this works, it can be a tremendous saver of manual effort. Think of the book as perhaps advocating an idiot savant vantage, and seeing how far one can usefully take this approach.

The results of the methods can also be used as input to more advanced and specialised methods, that rely on semantic analysis.

The book can also be applied to search engine analysis.