Bayes tages it worth 5.4 points. My limit is 5. How does Byes tag it this much? What should I do to solve the issue? Raise the limit to 7.5 or something would probably help on this one, but not all. A whitelist service is already created, but I can't keep track of 100 mails each day - and they do receive a lot of mail from different domains indeed...
Sounds like your bayes is heavily mis-trained... Personally, I'd just wipe the database and start over. This time take extra care make sure no legitamate mails wind up in the spam training, and make sure your ham training is representative of your normal mail.
I'd also push up the bayes spam autolearning threshold, to reduce the chance of a false positive nonspam mail from being autolearned as spam.
If you really want to know which tokens the email matches, run it through the command-line tool with debugging enabled.
spamassasssin -D < mistaged_email
You can also examine your bayes database contents just to get a general look at them.
sa-learn --dump
I like to redirect the dump to a file for easier viewing. Also when I am looking at general stats, I use head on the file. If I want to examine spam/nonspam tokens, I run the whole file through sort. Using sort winds up causing the header info to get shuffled in somewhere other than the top, but it also sorts all the tokens from nonspam to spam probability (0.000= strongly nonspam, 0.9999 = strongly spam)
------------------------------------------------------- This SF.net email sponsored by: Enterprise Linux Forum Conference & Expo The Event For Linux Datacenter Solutions & Strategies in The Enterprise Linux in the Boardroom; in the Front Office; & in the Server Room http://www.enterpriselinuxforum.com _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
