As part of my ML learning project, I decided to implement Paul Graham's implementation of bayesian filtering into python.

To run anything, fill "bad.txt" with something and "good.txt" with something else, and run "naive_bayes_prob(str)" on something else, and this will predict its "badness" (just a label, not anything else)

If you want to play with this, here are some ideas:

• use web.py to make a web based filter or predictor
• use this as a filter on your email to find "importance"
• use this on RSS feeds to alert you of posts you like