Making better spam killers
Paul Graham writes interesting stuff again. He’s done a follow up to his Plan for Spam article, detailing how to improve a bayesian system by adding contextual values — differing of words whether they’re in the header or in the body of the email, for example.
Most interesting. His Plan for Spam was what made me look at bayesian logic and switch from SpamAssassin to SpamProbe.
Unfortunately, I never get any spam at all to my own domain, so I haven’t had much use for it so far.