Interesting, but not exactly surprising. I guess when your source data is a whitelist of adulations, there are only so many conclusions you can make. NLTK looks interesting though.