The NLTK python lib is a great tool. I've been playing around with this recently and having a lot of fun. More recently I started looking at SCOTUS oral transcripts:

http://jarv.org/2012/02/a-text-analysis-of-supreme-court-ora...

The two articles here by the same author are interesting for folks who are interested in more about sentiment analysis

http://streamhacker.com/2010/05/10/text-classification-senti... http://streamhacker.com/2010/06/16/text-classification-senti...