[Spambayes] SpamBayes Outlook Plug-In Not Performing Well
Richie Hindle
richie at entrian.com
Tue Nov 25 17:48:36 EST 2003
Harri,
> What is the best way to retrain the database? It seems that SB already
> has all that it needs:
>
> pop3proxy-ham-cache (91 msgs)
> pop3proxy-spam-cache (495 msgs)
> pop3proxy-unknown-cache (21 msgs)
You're right, and retraining is exactly the reason that trained messages
are kept like that. However, no-one every got around to implementing the
feature. Retraining from scratch is so easy - you only need a few dozen
messages to get good results - that there's been no pressure for the
feature. And of course the developers (that's mostly me in the case of
the web training interface 8-) know to keep their database in balance, so
they don't run into the problem.
The next feature release has a feature whereby the web interface warns you
if it thinks your database is getting too imbalanced. That may prove to
be enough to avoid the problem.
--
Richie Hindle
richie at entrian.com
More information about the Spambayes
mailing list