[Spambayes] Basic Stuff

Tony Meyer tameyer at ihug.co.nz
Thu Sep 30 04:27:49 CEST 2004

[Philip Delancy]
> Do I have to keep training all the messages that
> keep coming in?

No.  In fact, training on every message is generally not the best thing to
do.  As Amir indicated, just training on unsures (any any mistakes, if there
are any) is probably the best thing to do.  (But the golden rule is that if
what you're doing works for you, then don't change it).

There's a whole heap of information about training on the wiki:


> What happens to the messages if I don't train them?

After a configurable amount of time, they are discarded (this doesn't effect
the copy that ends up in your mailer, of course), just like if you selected
the 'discard' option for each of them.  By default, this is 7 days, but you
can change that if you'd like.

> Is there anything else I should be doing, with the
> Clues/Tokens thing for example?

The Clues/Tokens links are only there for interest or trying to figure out
scoring problems (eg "why do these spam messages keep getting scored as
ham?").  They have no training functionality.

[Amir Katz]
> However, maybe (and I'm sure the developers can answer that)
> when a message that was categorized as spam has the radio
> button set to 'spam', it has the same effect as 'discard'.

No.  SpamBayes doesn't do any training unless it's told to (i.e. messages
that are classified as spam, aren't automatically trained as spam).
(Although there is an off-by-default option to turn this on, IIRC).  This
would probably be a bad idea (see the wiki stuff linked above).

=Tony Meyer

Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes. This
way, you get everyone's help, and avoid a lack of replies when I'm busy.

More information about the Spambayes mailing list