Consider the following spamassasin setup: auto_learn_threshold_nonspam 4.99 auto_learn_threshold_spam 5.0
(that is: auto-learn from *every* message) Messages identified as spam are saved (by procmail) in a "spam" mbox. All others go to the inbox. Whenever I see (in my inbox) a message that is spam (false-negative), instead of deleting it, I save it in the "spam" folder, along with the other (automatically classified) spams. >From time to time (more often in the beginning) I manually scan the "spam" mbox, and if I see a message that is ham I save it in a "ham" mbox, and delete it from "spam". Once every couple of days (and less often after some time) I re-train the system using the contents of the "ham" and "spam" mboxes. Would this approach give exactly the same results as if I was manually training the system? Am I right to assume that the undesired effects of learning from a false-negative (or a false-positive) will be undone when the system is retrained from the "ham" and "spam" mboxes as descrived above? Will the messages in "spam" that have allready been "auto-learned", be ignored when re-training the filter (as they should)? When re-training the filter using the "spam" and "ham" mboxes, will the headers added by SpamAssassin, be by default ignored? Thanks in advance /npat -- The human mind ordinarily operates at only ten percent of its capacity---the rest is overhead for the operating system. ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
