Note I said "raw"; by that I meant "before any filtering". Also, I was speaking about manual training, though I could see where autolearn might lead to the above ratio.
Has anyone seen good results from autolearn? I never did so disabled it long ago. Have seen many (other's) Bayes DBs become corrupt from autolearn / AWL...
IME you'll have better results by manually processing only a small amount of representative ham and spam through sa-learn. By small I do mean only a few per week. Our Bayes Tips are online at: http://www.postconf.com/docs/bayes_tips.shtml Roger Marquis
