Evidently I did something wrong when I installed the release version of 3.2 over earlier betas as this mailer is my first experience with Baysean filtering. It was about 95% accuracy with betas. When I set up the release, I wound up having to start the filter learning process again. I now have 37,000 bad and 17,000 good "words" in my filter. I suspect I screwed up something somewhere as I'm on getting 15% accuracy on junk mail.
How do I delete the current "word" lists (good and bad) so I can start the baysean filter in the learning mode again at 0/0 words?
Thanks in advance,
Shy
[Edit by Robin: moved to the Junk Mail forum]
[Edit by Eric: corrected subject]