On Tue, May 25, 2004 at 11:50:02AM -0400, Chris Santerre wrote: > > >For spam you are indeed adding to the DB for nothing. But Bayes is also > >about ham signs - I'd have thought that the From headers of > >your regular > >correspondents would be pretty good ham signs. > > Wait for it.......wait.......AWL! Ah, now don't you feel better now? ;)
It's not the same thing though. Bayes learns on any component of the ..From: header (escape _this_ line, sendmail !). So I get tokens for all the components of the address. AWL only works on the [EMAIL PROTECTED] part of it. AWL is also absolute, while Bayes can score on partial domains, for instance H*F:D*biz is a strong Bayes spamsign here ! > World Peace and Puppies, Awww innit cute - want one of our cats to go with them ? :) (thinks ... if he can solve world peace, he can surely get cats and dogs to live together !) Nick
