Re: [gentoo-dev] [ANNOUNCE] bayesiam SPAM filter 'bogofilter' ebuild on bugzilla

[Chad Huneycutt]

On Fri, 2002-11-08 at 15:22, Alan wrote:
> The two limitations of bogofilter that need to be addressed someone are:
>  1 can only "interact" with it through mutt or some custom hacking
>    through other programs (I think that emacs vm can notify
>    spam/non-spam with bogofilter if you know the right spell)
>  2 needs enough spam and non-spam to properly set up the word lists.
>
> However, if 2 is done right, then 1 isn't a real problem.

These are excellent points.  I saved lots of stuff from my Yahoo! and
Hotmail bulk folder, as well as other spams I got for about a week to
train bogofilter.  I don't remember how much of my good mail I trained
it with, but I have a lot saved.  Finally, once I installed it into my
mail chain via procmail, I used fetchyahoo to get mail from my Yahoo!
account from my inbox and Bulk mail folder, so I could monitor the
results for a while.  No problems whatsoever.  Once I was convinced that
the Bulk mail folder stuff wasn't getting into my inbox, I stopped
retrieving that.

As far as point 2, what I did is to create a folder that I put spam in,
if it happens to get through.  I then have a cron job that runs the
contents of that folder through bogofilter -S (this is not ham, this is
spam!).  I don't have the folder for fixing false positives, but then
again, I haven't had any.

If you are interested, here is the script that I run from cron every few
hours to clean out my spam folder.  Note that there is a locking issue
with my rm and touch of the spam file, but I didn't really care:

#!/bin/sh

$HOME/pub/bin/linux/bogofilter -S < $HOME/mail/Spam
/bin/rm -f $HOME/mail/Spam
touch $HOME/mail/Spam

Chad Huneycutt



--
gentoo-dev@g.o mailing list