Naive Bayes classifier

I recently did a project with naive bayesian classification, and it was a blast. If you haven't gotten your feet wet with artificial intelligence yet, I recommend document classification as a starting point.

O'Reilly's "Collective Intelligence" has a great chapter on this, with at least one improvement over the stuff I culled from the rest of the web. Paul Graham has a post somewhere about spam filters that shows how the big kids build it.

For the kids in the front row, try changing your old classifiers to 4-step markov chains. If you're asking "why would I do that?", you're in the wrong classroom :)