forum wrote: > I am a student and testing SA for my project. > Thus, I change sub tokenize_line to produce an output that i want. Now > my Bayes was learnt about 350 hams and 350 spams but it has only 650 tokens. > > When I use spamassassin -t command to test my mail, Bayes was working > properly. It give a spam score as i desire. When I try to send mails via > outlook express, there are no score form Bayes. > > Why it don't produce any score? > Is it because I've go too low amout of token?
Yes. I'm not sure how you got so few tokens from 700 emails, but SA requires a decent number of tokens (2,000?) before it starts scanning. I'd expect if you trained 700 messages you'd have at least 3,000 tokens, and probably much more than that. I think you might have broken tokenize_line. Try piping a message into spamassassin with debug output and you'll see some information about why bayes is inactive. spamassassin -D <sample-nonspam.txt