[Spambayes] Re: [spambayes-dev] wish: autotraining

Tony Meyer tameyer at ihug.co.nz
Mon Nov 8 03:05:03 CET 2004

> The setting I am referring to is in the advanced settings
> page. 

Sorry - I didn't realise that we exposed that option.  I've checked in a
change to the text to make it clear that it doesn't apply to the POP3 proxy.

> I thought it would mean that the autotrained spam-messages
> would not appear in the (manual) traininglist marked as spam.

It's not a particularly suitable solution for the POP3 proxy for a couple of

 1. Mis-training a message has very negative consequences for future
accuracy.  This means that if the proxy trained based on the classification,
and you didn't correct the mistake, classification accuracy would get
progressively worse.

 2. This would (without other modifications) mean that you were using "train
on everything".  This is not the recommended method of training, and doesn't
lead to optimal results.  <http://entrian.com/sbwiki/TrainingIdeas> has more
information about training techniques.

> Do I read that it wont work because this is not a
> "procmail-based solution" ?

Yes, only the sb_filter.py script makes use of this option.


Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.

More information about the Spambayes mailing list