On Sun, July 22, 2007 14:41, S James S Stapleton wrote: > Background; I've been using Spambayes for a while, it starts out good, > quickly > is trained to great, but at some critical point drops to crud. > > At the great state, I get no false positives, and the false > negatives/uncertains are low. But spammers change their tactics and I have > too keep training to keep SB in the great state, until at one point it > gets > overtrained and *everything* pretty much goes into spam and I just get the > few (1-5%) false negatives, and some mailing list stuff, in my box. > > (1) Is there any way to whitelist email addresses? The only address I get > spam > from regularly is when they spoof mine. > > (2) Can values be modified and locked? For example, I'm part of several > mailing lists which tag their emails, [PyQt], [PHP], [PHP-WIN], [SDL], > etc. I > would like to be able to mark those tags, and some common keywords things > like that as 'minimal spam probability]. Add to that some keywords I know > I'll never see in a legitimate email, I wouldn't mind fast-fowarding them > to > the 'this is a spam word' state. > > Please CC me any responses, as I am not on the list.
I learned from your mail headers that you use Kmail so there is a 99% possibility you're on Linux. Is that correct? How do you have SpamBayes set up? I use it as a procmail filter, like this: (I don't have access to my .procmailrc ATM, perhaps later) * First, a few rules that move mailing list traffic & mail from friends to certain folders. This is a sort of procmail whitelist. These kinds of mail don't get spamfiltered, however I have used them for initial training. --> This already answers your two questions. * A rule that scores mail with SB * Everything with spamscore 100% gets the /dev/null treatment * Remaining spam is moved to my spam suspect folder. (to move manually to the training folder later) * Unsures are _copied_ to my unsures folder * Ham is moved to my inbox or processed by other procmail rules. The exact syntax of this .procmailrc file is left as an exercise to the reader. :) -- Amedee _______________________________________________ SpamBayes@python.org http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
