This is very cool work! Some years ago, I was interested in text mining. I ended up playing with latent semantic analysis using Lucene etc. But that was a largely random choice, driven by the availability of open-source software and online discussion.

However, as cool as stylistic analysis is, I'm concerned about implications for online anonymity (which I consider valuable). But maybe the risk is limited by typical text length and false positive rate. I welcome suggestions for further reading.