Mining the Social Web, by Matthew A. Russell
This book covers a lot of ground. It's, at times, a bit vertiginous in the amount of subjects and technologies it touches per chapter, and is not always easy to follow. It can also introduce so many interesting things that, by the time you finished becoming familiar with all of them, after wandering for hours on the web, jumping from interesting technology to interesting technology, you may have forgotten what took you to these places and wonder where you were in the book. Time spent reading it is, however, time very well spent. When you finish it, you will have at least a cursory familiarity with tools like OAuth, CouchDB, Redis, MapReduce, NumPy (and the Python programming language, albeit it will help you a lot if you know your way around Python before you start the book), Graphviz, SIMILE widgets, NLTK, various service APIs and data formats, and will be well equipped to explore those rich datasets on your own. The chapters are well compartmentalized and it's easy to pick chapters to read according to your needs. I know that, when I face the problems they tackle, I will do exactly that.
If you do any kind of analysis and visualization of social-generated data that's on the web, this book is a good pick. Even if your datasets are not from the web, you may find the parts on analysis and visualization very interesting.
Disclosure: I reviewed this book for the O'Reilly Blogger Review Program. If you have a blog and love to read, you should take a look into it. It's fun.