Joel Uckelman on 5 Nov 2003 05:08:41 -0000

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

[hosers-announce] bogofilter upgrade

I just upgraded bogofilter from an ancient 0.13 build to a quite recent 
0.15 build. Included in this newer version is improved tokenization of 
mail---bogofilter now distinguishes between tokens appearing in the headers 
and tokens appearing in message bodies.

In order to take advantage of this, you need to rebuild your goodlist.db 
and spamlist.db. You can keep using your old db's, but new messages will be 
added with header/body classification, and having mixed data might hurt 
classification performance.

So, if you still have all the mail you used to train bogofilter, just 
remove your current databases and use mh2bogo to do the retraining. If you 
don't have any of that mail still around, save all of your mail for a few 
weeks and use that. The spam and good databases should contain about the 
same number of messages; I can supply any number of spams up to about 8100 
if you need more to make up the difference.

If any of this is confusing, please ask.


hosers-announce mailing list