Hi ,
i want to configure mail server which can filter spam using Bayesian
classifier. I start experimenting with SpamAssassin and i am not sure
how everything should look like. As much as i understand the user must
specify if the mail is spam or ham, until he create appropriate
database. How have u didi it , can u give me example . Thanks for any
help , because i am in the middle of nowhere

[ Post a follow-up to this message ]

On Tue, 13 Jan 2004 22:24:03 +0100, Stefaan A Eeckels <tengo@DELETEMEecc.lu>
poured into the usenet group comp.unix.admin:
quote:

> On 13 Jan 2004 08:03:50 -0800
> vangelob@in.tum.de (murph) wrote:
> 
Bayesian filtering is user specific. You might find setting up spamassassin
easier.
[QUOTE] 
>
> Have a look at Eric Raymond's bogofilter. You can prime it
> with a body of spam and ham. It works really well, with
> nary a false positive and an excellent tag rate of real
> spam.

Hmmm, on the server? For me, all text/html is spam. For you, it might not
be. On a server, I would rather reject mail than classify it. If you
classify mail, you have to look for false positives and that essentially
nullifies the advantages of classification.
Filtering really doesn't scale up.

Devdas Bhagat

[ Post a follow-up to this message ]