Re: bogofilter behavior

[Matthias Andree]

Harry Putnam <reader@newsguy.com> writes:

> This may be a little out of line for this groups.
>
> The bogofilter mail list seems to be pretty dead.  So please bare with
> me for moment.

Have you tried contacting the list owner or one of the developers?

BTW, I just allowed myself to send a test mail to the list, being one of
the bogofilter maintainers and list moderators, and it appeared on the
list in less than one minute.

Can you confirm your own mailing list is OK? Have you checked your
subscription? If your mail address has refused mail over several days in
the past, the list software may have put your subscription "on hold",
i. e. "nomail".

> I'm using bogofilter in procmail not gnus, but if someone is familiar
> please talk to me off list.
>
> I'm using the most primitive setup.  Where procmail runs bogofilter
> then checks the bogofilter header for spammyness.
>
> In bogo_spam.in group some false positives occur and I run them back
> thru bogofilter according to my reading of the man pages, this should
> correct any misfiling that occured in bogo data base.  
>
> `bogofilter -v -Sn < MSG_NUM (a false positive file name)

Use the -S only if you are using bogofilter (in procmailrc) with the -u
option, use bogofilter -vn <MSG_NUM if you are using bogofilter without
-u option in procmailrc.

However, if that message has lots of tokens that appeared in spam often,
it'll take a while for bogofilter catch up. The list archives have
several discussion about "training to exhaustion".

> However some of these false hits have recurred several times and are
> very similar.  I was under the impression that the above command would
> tell bogofilter these are not spam.  It seems to have no effect.

That may well be. If you've had a mail with 50 "good" and 100 "spam"
tokens, and each of the spam tokens appeared in 20 spam mails, it'll
take near 20 "good" mails to have these tokens before bogofilter has
learnt "I don't know if this token is good or bad, it appears in either
kind of mail".

Check bogofilter -vvv what it thinks and what token make that message
"spammy" - and what the counts were.

-- 
Matthias Andree

Encrypted mail welcome: my GnuPG key ID is 0x052E7D95 (PGP/MIME preferred)