> -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] Behalf Of Paul > Murphy
> The problem is that the documentation suggests that you > enable the two options, > and sit back and watch the database being built as e-mail comes in. > > The reality is that the two options enable the use of the > bayesian filter, but > the database remains empty until you have trained it on a > suitably large dataset > of spam and "ham". > > Does anyone have a way of using Mimedefang to automatically > build the database, > even if that means forwarding a copy of every message to > either a spam or ham > mailbox, and then processing and deleting the mailbox daily? you can use the auto_learn and avoid the training. You just have to be patient. very patient... eventually you start seeing "BAYES_" showing up in your test result. For me, that amounted to almost 2 weeks of patience. But it was worth it. The bayes score is helping a lot now. -lee
_______________________________________________ Visit http://www.mimedefang.org and http://www.canit.ca MIMEDefang mailing list [EMAIL PROTECTED] http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
