Incorrigible spam

[David Relson]

On Wed, 7 Apr 2004 11:11:27 +0100
Richard Kimber wrote:

> >From time to time I get spam that is still scored as ham after I've
> >told
> bogofilter to relearn it as spam.
> 
> What's the best strategy for these sort of messages? Should I just put
> up with this and just delete them, or is there an approved way of
> dealing with them (e.g. should I not train with them?)  I tried
> relearning the message four times, but the spamicity hardly changed
> (0.610188 to 0.617225).
> 
> - Richard

Hi Richard,

Sorry to say, but there's no consensus on the best way to train
bogofilter, much less on strategy for difficult messages.

I tend to train on everything, use tri-state scoring with the -u switch
(autoupdate) and additionally train with unsures and errors (false
positives and false negatives).  For me, no message is more or less
important than any other.

Another common practice is an inital train on everything (for several
thousand messages), with train on error after that.

Some people take misclassified messages and keep training with them
until bogofilter gets it right.

The above practices, and others, are all used and all allow bogofilter
to get its work done.  Picking a strategy and sticking with it seems to
be what's important.

Enjoy!

David