I would have mentioned http://www.paulgraham.com/spam.html earlier, but I personally don't feel the approach is where they say it is. I was getting way too many false positives when I tested CRM114 (which claims 99.87% accuracy) -- http://www.paulgraham.com/wsy.html
> Not possible? ;) > http://www.google.com/search?hl=sv&ie=UTF-8&oe=UTF-8&q=bayesian+filterin > g&lr=
