Bayesian database not learning
Douglas Ward
dward at nccumc.org
Tue Oct 10 17:09:41 IST 2006
Thank you for this information. I checked /root/.spamassassin and saw
a 1 kb bayes database. I assume this is the blank database created by
spamassassin upon installation. My real bayes database is over 2.5
mb. Since I am not using the blank bayes db I deleted it. I then ran
the following commands:
sa-learn --dump magic
sa-learn --dump magic -p /etc/MailScanner/spam.assassin.prefs.conf
Both returned the same number of spam/ham tokens. I think the problem
spam wise is that my learn value (12.0) is too high. I am curious
about the low end value (0.1). Does this catch negative scores? Most
of our ham scores less than zero but it is not learned either. The
message that I received this morning with a score of 16.5 surely
should have trained the tokens.
Time for the potentially silly question: What value should
bayes_auto_learn have (0 or 1)? How about bayes_auto_expire?
On 10/10/06, Billy A. Pumphrey <bpumphrey at woodmclaw.com> wrote:
> > -----Original Message-----
> > From: mailscanner-bounces at lists.mailscanner.info [mailto:mailscanner-
> > bounces at lists.mailscanner.info] On Behalf Of Douglas Ward
> > Sent: Tuesday, October 10, 2006 10:21 AM
> > To: MailScanner discussion
> > Subject: Re: Bayesian database not learning
> >
> > I did check that. Wouldn't sa-learn know which database it was using
> to
> > learn?
> >
>
> By default it will use a different database, /root/.spamassassin/ I
> believe. From my experience, you have to specify the -p for the conf
> file. I tested this by doing a sa-learn without it and then checked the
> bayes database with the dump command and there were no new updates. I
> added the -p command and the bayes was updated. Looks like you may not
> have to add the -p to the sa-learn dump command though, although you may
> want to make sure yourself.
>
> Here are my notes from when I set my learning up:
>
> sa-learn --dump magic
> This will show you how many emails bayes has learned
> http://www.annodex.net/cgi-bin/man/man2html?sa-learn+1
> Good link
>
> HERE IS WHAT TO USE for the spam from exchange to linux learn:
> As the spam user:
> For spam
> 1. fetchmail --folder spam --all (logged in as spam)
> 2. Log in as root
> 3. sa-learn -p /etc/MailScanner/spam.assassin.prefs.conf--mbox
> ---no-sync --showdots --spam /var/spool/mail/spam
> 4. rm -f /var/spool/mail/spam
> 5. touch /var/spool/mail/spam
> 6. sa-learn -p /etc/MailScanner/spam.assassin.prefs.conf --sync
>
> For ham
> 1. fetchmail--folder ham --all (logged in as spam)
> 2. Log in as root
> 3. sa-learn -p /etc/MailScanner/spam.assassin.prefs.conf --mbox
> --no-sync --showdots --ham /var/spool/mail/spam
> 4. rm -f /var/spool/mail/spam
> 5. touch /var/spool/mail/spam
> 6. sa-learn -p /etc/MailScanner/spam.assassin.prefs.conf --sync
>
> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
>
> --
> MailScanner mailing list
> mailscanner at lists.mailscanner.info
> http://lists.mailscanner.info/mailman/listinfo/mailscanner
>
> Before posting, read http://wiki.mailscanner.info/posting
>
> Support MailScanner development - buy the book off the website!
>
More information about the MailScanner
mailing list