Advice / Resource on setting up Bayas

[Brian McBride <br...@vervelab.com>]

Hello everyone,

I am using SA with MailScanner on Redhat 7.x using virtual email boxes.

Currently, I forward all mail that gets a high spam rating from MailScanner
to a spam mailbox.  MailScanner is set to add to the header of the email
with the scoring that it has found, but other than that, it seems that the
headers are in tact (the is no forwarding header to spam@domain.com, just
the user@domain.com ).

MailScanner also modifies the subject line.

Last, my spam mailbox is site wide, encompassing 3 domains.  I am dumping
all spam in one area (mostly for simplicity sake).

There is some spam that does not get filtered, or is on the low spam score
so I have that come into my mailbox.  I can leave my messages on the server,
but using POP, they all just sit in a big file.  I use outlook to check my
various email account boxes in one location.


Here are my questions:  Can I get an accurate training using messages that
MailScanner has modified, or will those headers mess everything up?  Does
anyone know of any plugins to Outlook that might allow exporting multiple
emails into a mailfile that SA can learn from?  That would be so handy as I
would not have to sort my mail on the server as well as locally.  Last, as
my spam folder is site wide, will the fact that emails going to
user1@domain.com, user2@domain.com, etc... cause problems since my good mail
file would only contain the emails with the destination me@domain.com?


Bayas sounds great, but if I find that I spend the same or more time
filtering my email, then I feel rather silly using a program like
SpamAssassin.  What I want, like us all, is to never see the spam messages
and not have to waste my time with them.


Thanks in advance for helping out a SA newb.

Brian!

Re: Advice / Resource on setting up Bayas

[Kai Schaetzl <ma...@conactive.com>]

Brian McBride wrote on Tue, 13 Apr 2004 15:50:50 -0700:

> Bayas sounds great, but if I find that I spend the same or more time
> filtering my email, then I feel rather silly using a program like
> SpamAssassin.  What I want, like us all, is to never see the spam messages
> and not have to waste my time with them.
>

That's what autolearning is for. Train Bayes until it works nicely then rely 
on autolearning and only train from time to time if you have a pile of spam 
easily available at hand, f.i. in a spamtrap account. I haven't trained our 
Bayes for months but it scores on almost every spam message.


Kai

-- 

Kai Sch�tzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org