New user and question
David Relson
relson at osagesoftware.com
Sat Oct 30 20:15:55 CEST 2010
On Sat, 30 Oct 2010 14:10:54 +0200
Lars Clausen wrote:
>
> On Oct 29, 2010, at 1:36 PM, David Relson wrote:
> >
> > Continue to train with all the unsures. That's pretty much
> > necessary to keep bogofilter up-to-date.
> >
> > Be careful with "-u" and false positives (FP) and false negatives
> > (FN).
> > When using "-u" false results tend to magnify unless you watch for
> > such
> > message and retrain properly with them. For a FP (ham scored as
> > spam),
> > it is necessary to run bogofilter with the -S and -n flags. For a
> > FN (spam scored as ham) run bogofilter with the -N and -s flags.
> > These flags are described in bogofilter's man page.
>
> Is it any problem to run unsures through with -Sn or -nS? I just
> have two IMAP folders for mails I want to retrain and some cron jobs
> to train from there and move them out. I would rather not have to
> complicate it with splitting into separate folders for wrongly
> classified and unsures as well.
>
> -Lars
Hi Lars,
Good question. It's OK to use -Sn and -nS (as appropriate). If you
want to see the ham/spam counts and scores for a token, use bogoutil.
As an example:
$$$ bogoutil -p wordlist.db Lars
spam good Fisher
Lars 138 121 0.073639
You can run the command before and after traing with a message and
you'll see the change in the spam and good counts and in the Fisher
(a.k.a. "bogosity") score.
HTH,
David
More information about the Bogofilter
mailing list