Have a look at ASSP, a PERL based anti spam Bayesian pre filter and run by a web interface. It's very solid and accurate, easily trained. I ran it for several years with only 2 restarts. It took about ten days to get to around 99.5+% accuracy provided you did the work to train it. Look for it at Sourceforge.
Authored Comments
Never realised this existed till just now. Thanks for the article.
Have a look at ASSP, a PERL based anti spam Bayesian pre filter and run by a web interface. It's very solid and accurate, easily trained. I ran it for several years with only 2 restarts. It took about ten days to get to around 99.5+% accuracy provided you did the work to train it. Look for it at Sourceforge.