On Tue, March 31, 2009 12:33, Arvid Ephraim Picciani wrote: > what exactly is the point of learning > specific messages when the next one will be different anyway.
for bayes its still count to learn, not just spam, the more learned tokens, then bayes works better, try get it learn with equal ham/spam tokens not just 100000 spam mails and 200 ham mails, see the numbers from sa-learn --dump magic if to much learn on spam / ham addjust the autolearnthreashhold if set to around 5 it will be like in a ship that sail in a storm :) -- http://localhost/ 100% uptime and 100% mirrored :)
