Is it possible for sa-learn to learn too much? I have an old users mailbox that we have basically been using as a spam trap for the last year or so. There are 6000 messages in it right now and a lot are dupes. Do duplicate messages cause the learn process any problems? If I dedupe it I get about 4500 messages. If I learn all 4500 of those messages can I start getting more FPs?
I can already tell that it is learning because it is starting to show the BAYES_90 in the reports and our tagged spam numbers are increasing. Should I run all of them through sa-learn and then if users get FPs have it unlearn those and/or have it learn them as ham? -Mike Carlson [EMAIL PROTECTED] http://www.uselessthoughts.com ------------------------------------------------------- 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
