On Wed, April 30, 2008 02:09, alan falk wrote:

If you can describe the process [which is what they did,] it can be programmed. That's all I meant.

Just as SpamBayes filters things which I could be making a decision on, too...

There is a small but critical flaw in your reasoning. It is virtually impossible to describe the mental processes going on in my mind (and your mind) to determine if an email is ham or spam. If there ever were such an algorithm, there would be no need to train Spambayes. Heck, there wouldn't even be a need for Spambayes in the first place, because your theoretical algorithm would be the perfect spamfilter all by itself. Because trying to replicate human mental processes is doomed to fail, various approximations are used. A statistical approximation is one approach, and Spambayes is one implementation of a statistical approach to attempt to mimic human mental processes. It's not perfect, and it will never be, but it's good enough. -- Amedee