Thank you Matt =). So most of the heuristics seem to be looking for SPAM. What are the ones that would push a mail towards being HAM (and that are not ignored by autolearn bayes)? So far I have found one: ALL_TRUSTED.
Thanks! Breena On Wed, 26 Jan 2005 13:41:12 -0500, Matt Kettler wrote > At 05:37 PM 1/25/2005, breena wrote: > >I have been checking my logs and it seems that even when an email comes in > >with a whitelisted address (making its score below the default threshold for > >autolearn ham), it is not learned as ham. According to the logs, nothing is > >ever autolearned as ham. Does autolearn ignore the whitelisted mails? > > Autolearn ignores white and blacklist statements. The autolearner > evaluates based on the score of the message with bayes disabled and > all "noautolearn" rules disabled (ie: white/blacklists, AWL, and > GTUBE). > > It might still autolearn a whitelisted message, but it won't learn > it on the basis of being whitelisted or not. > > >What would have to be true about the email for it to be autolearned as ham? > > Here you go: > > 1) bayes_auto_learn must not be set to 0 > > 2) The pre-bayes/noautolearn corrected score of the message must be > below bayes_auto_learn_threshold_nonspam (defaults to 0.1) > > 3) The existing bayes rules (actually those with the "learn" tflag) > must not total to a score over +1.0 for the message as it is now. > This creates the "don't auto-learn as ham anything you previously > thought to be spam" rule. > > 4) the bayes DB write lock needs to be available (ie: nobody else > can be writing at this time). -- SPAM and Webmail solutions from CompanyV.com
