on Wed, Nov 12, 2003 at 10:19:01PM +0100, Benedict Verheyen (linux4bene@pandora.be) wrote:
> Hi,
>
> my Spamassassin's bayes stuff finally kicked it as i now see bayes_00
> and similar stuff in the headers. Do i need to keep feeding spam and
> ham to sa-learn?
Not necessarially (to contradict everyone else's advice)...but it
doesn't hurt.
SA has an "autolearn" feature, where mail scoring above 6, and below
0.1, will be "autolearned" as spam and ham. That is, the Baysian
classifier will train on these mails.
I've set up a mailbox "spam-learn" where I dump any spam which slips
past SA, and a cron job to run 'sa-learn' over this every 30 minutes.
While I could set up equivalent 'ham-learn' and 'sa-forget' folders, I
haven't had the need as yet. The effect is that I can run spamassassin,
auto-train off most of the mail and explicitly train based on any false
positives or negatives which slip through. Overall effectveness is
quite high. Design is simple.
Peace.
--
Karsten M. Self <kmself@ix.netcom.com> http://kmself.home.netcom.com/
What Part of "Gestalt" don't you understand?
Bush/Cheney '04: Four More Wars!
Attachment:
pgpKSyi7n_fZS.pgp
Description: PGP signature