[warning: 99% OT] does anything like this exist ?...

Andrew Dalke adalke at mindspring.com
Wed Oct 27 04:02:50 CEST 2004

Fred Pacquier wrote:
> * is the documented change on line 348 enough to run the script with a  
> current version of Spambayes ?

I upgraded to Spambayes 1.0.  I think I needed to change something, but
I've forgotten.  I've attached my current version (sans passwords and
identifing mailing list information) to this post.

> * does use of sb_culler contribute to the training of the Spambayes db, 
> or does it assume that it is kept current independently (by means of 
> normal use by a mail client through the POP3 proxy for instance) ?

It does not contribute.  It can't.  That is, it computes a spam value
for a message but since it doesn't know if that classification is
correct it can't train itself.

Every few months I export all my saved mail (inbox, sent, various
archive folders) as "ham".  I also save all the spam that gets through
my filter so I can export it as "spam."  I then train SpamBayes on
those two data sets and use the result for my filter program.

> Well, if any of these new additions are available for outside use, I'd 
> sure appreciate a copy of the updated script...

I didn't make those additions to the attached file.  Feel free to do
so, and to update the Wiki.  Shouldn't be more than a dozen or two
lines of code.  Eg, to have a file of blacklisted subjects you could
use WhiteListFrom (which reads lines from a file, and reloads if the
file's timestamp changed) and tweak the test condition.

Enjoy, and thank you for your kind comments regarding my code.  That
program has really helped with my email sanity.  Thanks also to the
SpamBayes people and to Kevin Altis for his virus killer code I used.

				dalke at dalkescientific.com

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: sb_culler.py
URL: <http://mail.python.org/pipermail/python-list/attachments/20041027/0abd3774/attachment.ksh>

More information about the Python-list mailing list