On Sat, 29 Nov 2003, Gorm Jensen wrote: > Bayes is working very well for me, but I am concerned about poisoning the > database with extraneous, obfuscating words that many spam messages contain. > > A few postings to this list say that there is no problem, but I don't want > to spoil a good thing. Are there some rules of thumb on what to sa-learn > and what to avoid?
I had my bayes db poisoned recently. I think it was because of the ratio of spam to non-spam that I get: 10-20 legit emails a week versus about 100+ spams a day. I'm trying a different strategy now to even the ratio. I call SA first without bayes and then filter anything scored above 15. Then I later call SA with bayes like I normally would. The first pass at 15 catches about 75% of my spam without any false positives. I don't feed this into the bayes db -- only the ones caught/not caught in the second pass. ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk