Bootstrapping the database in a MUA-embedded scenario
Mikhail Zabaluev
mhz at altlinux.org
Wed Aug 10 23:49:39 CEST 2005
В Срд, 10/08/2005 в 17:29 -0400, David Relson пишет:
> > > > bogofilter -n -H </dev/null
> > >
> > > That works too!
> >
> > Nope, it doesn't update the word count. Too bad.
>
> If the wordlist exists, bogofilter _can_ score messages. The following
> commands demonstrate this:
>
> rm -f wordlist.db
> bogofilter -d . -n < /dev/null
> bogoutil -d wordlist.db
> echo this is a test | bogofilter -vvv -d . -H
>
> Of course bogofilter is happier when there's at least one ham and one
> spam message
That was my point: given no hams to learn, it scores 0.52 no matter how
many spams it has learned. To receive a ham message for learning from
Evolution, it needs to classify something as spam first.
More information about the Bogofilter
mailing list