It seems to me that the autolearn feature is too withstrictive as to which mails it chooses to autolearn from, where as my plan would autolean from _all_ the mail I receive, and only require I keep and eye out for any mistakes, and correct them.
If you're simply concerned that the autolearn thresholds are too restrictive, you can always just change the thresholds to more aggressive numbers... The only big restriction is that to learn as spam it always requires 3 header and 3 body points... But you can lower the bayes_auto_learn_threshold_spam from 12 to 6 and increase the number of emails SA autolearns quite a bit. You can also raise bayes_auto_learn_threshold_nonspam to a number higher than the default 0.1.
This would result in fewer mis-learns that would need to be hand corrected compared to learning everything. You'd still have some, but it would be fewer to fix.
The biggest drawback of the "learn everything and hand fix the mistakes" is that unless you hand fix in more-or-less realtime, your mis-learning is going to cause related emails to be misclassified until you re-learn them.. This could be a continuous uphill battle on any account with a decent mixture of spam and nonspam.
------------------------------------------------------- This SF. Net email is sponsored by: GoToMyPC GoToMyPC is the fast, easy and secure way to access your computer from any Web browser or wireless device. Click here to Try it Free! https://www.gotomypc.com/tr/OSDN/AW/Q4_2003/t/g22lp?Target=mm/g22lp.tmpl _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
