Spam filtering.

A few days ago, I opened my ~/.mutt/muttrc and added the following:

macro index,pager S s=spam/gemist<enter><enter>

=spam/gemist is my folder for, well, missed spam. False negatives. Mails in that folder get read by a cronjob that feeds them to sa-learn --spam, and I found I was having way too much of those lately. I was already tagging them per mailbox, so that I could throw them to the false negatives box all at a time, but these were still way too many characters. Hence, the macro to throw spam away with only one keystroke (well, two if you count shift).

While doing this and looking at my mails more closely than normal, I found out why I was having too much of those, too: the score of the BAYES_99 test (meaning, the bayesian filter was 99 to 100% sure that the mail was spam) was set to 1.9! Freakin' idiot. Modified it in my local configuration to say 4.0 rather than 1.9, and suddenly a lot more spam was caught.

Happy happy, joy joy.

Not sure whether this is my bug or whether it's in the package; I'll have to review that.