[math-fun] Spam-blocking, Bayesian methods

I would appreciate a pointer to a readily accessible article or three on how Bayesian methods can be used to block Spam. Pointers to any products implementing these methods also welcomed. Guy