OK - I'll start feeding it the a couple of my most common folders. Let's see if it helps.
As for headers, I'm using the mail saved by pine, and I think that saves everything. I still think its not a good idea to have the BAYES_99 score set higher than the cutoff. Too easy to generate false positives. But let's see how things go. On Wed, 1 Oct 2003, Tom Meunier wrote: > 1. It's okay to train with the spam with Spamassassin markup. sa-learn > ignores the markup. > 2. It's not okay to train with "this month's sent-mail" - you need to > be training it with mail that reflects the type of mail you GET, not > send. The headers are as crucial as the body. > > -tom > > > -----Original Message----- > > From: Jack Gostl [mailto:[EMAIL PROTECTED] > > Sent: Wednesday, October 01, 2003 7:04 AM > > To: Matt Kettler > > Cc: SpamAssassin listserve > > Subject: Re: [SAtalk] Bayes (again) > > > > > > By the way, when I say "caught", I'm passing in the spam, > > heaer and all, that SpamAssassin caught. Maybe this is > > causing the problem. > > > > > > > > > > I did the initial training with my collection of "caught" spam and > > > with my current month's "sent-mail" plus one or two other > > folders of saved ham. > > > > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Spamassassin-talk mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/spamassassin-talk > -- Jack Gostl [EMAIL PROTECTED] ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
