On Wednesday 31 December 2003 20:59, Gil Freund wrote:
I wonder, does bayesian filtering make sense on a domain level (i.e. the same DB for all users) and not having each user teach the system his/her own rules?
Good question. I have no idea :-)
I've set it up anyway, and it looks to be working OK (that is no complaints from users so far :-). I know its not nice to do, but I occasionally scan user's inboxes by grepping for known keywords to extract SPAM that they got and then feeds it to the dictionary. I also have some dummy accounts which exist for the sole purpose of attracting SPAM.
How do you feed it? I thought SA reads MBOX and Maildir formats only?
All in all I think SPAM is generally the same for all the users - viagra ads and other suspect materials, nigerian scams and yambateva.
-- Oded
-- ========================================= Gil Freund Sysnet consulting ----------------------------------------- [EMAIL PROTECTED] http://www.sysnet.co.il voice: +972-52-676906 Fax: +972-8-9356026 =========================================
================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
