Basic Bogofilter Magic

21 October 2003

I've been using bogofilter to herd spam into it's own little mailbox for later deletion/review. My spam mailbox easily accumulated 300-500 messages a day. I need to recognize a false-positive in there, so the volume was still quite unmanageable.

I used to send everything with a 0.50 (of 1.00) rating to the spam box. Bogofilter actually has a third categorization of Unsure. I'm using this categorization now and have adjusted my thresholds. Bogofilter proved to be pretty good, so now I let it bounce anything over a 0.90, and then anything between 0.90 and 0.50 still gets saved to an actual mailbox. This has cut my volume considerably, so I have a better chance of catching mail that shouldn't be spam. It also keeps those border-line cases from skewing the auto-learn into producing more false-positives.

Filed Under: Computers Linux