-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello Joe,
Thursday, July 24, 2003, 7:43:02 AM, you wrote: JJ> I see what you're saying. I actually whitelisted these specific JJ> addresses due to the fact that all they produce is ham. Nothing that JJ> comes from them looks remotely spammish and I wanted bayes to learn JJ> from them (esp. to get over that 200 message hurdle). But, if nothing that comes from them looks remotely spammish, then where is the value in teaching this ham to Bayes? IMO the value of Bayes is that it learns to distinguish between questionable emails, those which more standardized rules can't pin down. Because of this, I concentrate on feeding Bayes those (very) few emails which are misclassified (FPs and FNs), and also those emails which are in the midrange (spam within 1.0 of my required hits, ham with positive scores). I believe this will tend to help Bayes more correctly classify the otherwise hard to classify emails. JJ> I suppose what I am looking for then is a new feature ie: "learn-from JJ> [EMAIL PROTECTED]". Or I suppose I can just make a rule that looks JJ> for the From that I want to learn from and set my own negative score JJ> for those. I considered doing something like that, creating a non-whitelist rule with a -50 score, and a non-blacklist rule with a +50 score, just to force Bayes to learn the email, but if the email is that easy to classify, then why bother filling up the Bayesian database with those tokens? Those emails are what the normal SA rules are for. YMMV. Bob Menschel -----BEGIN PGP SIGNATURE----- Version: PGP 8.0 iQA/AwUBPyCf45ebK8E4qh1HEQKCuwCfZ+9/U7Ou4SMKnnGa0zu0B9E5e60AoN2E jZ3MYSB4jjhCiUykOGW+Iv28 =Pv8q -----END PGP SIGNATURE----- ------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
