At Wed Oct 29 14:15:51 2003, Bill Polhemus wrote: > > Are there other criteria, though? > > For example, I have set the threshold at which Auto-Learn is > "triggered" for Spam at 7.99. Anything scoring over that is > designated to be "auto-learned." > > Yet one came through this morning at 12.9, and it did NOT > "auto-learn."
To avoid reinforcing its mistakes, auto-learning uses the non-Bayes sets of scores to decide whether or not to auto-learn any given message. However, the scores in the headers are from the with-Bayes scoresets. > However, Bayesian probability was something like 0.997, so I thought "maybe > at that level it figures it doesn't need to "auto-learn" this one. But when > I "hand-feed" it through SA-Learn, it accepts it! sa-learn will learn anything you feed to it (on the basis that you're a human and know whether a message is ham or spam), while the auto-learning is much more conservative to avoid learning spammy-looking ham in error. Martin -- Martin Radford | "Only wimps use tape backup: _real_ [EMAIL PROTECTED] | men just upload their important stuff -o) Registered Linux user #9257 | on ftp and let the rest of the world /\\ - see http://counter.li.org | mirror it ;)" - Linus Torvalds _\_V ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
