[gentoo-user] Improving SpamAssassin's accuracy...

[Stroller]

Hi there,

I emerged SpamAssasin on a mailserver the other day, added the  
appropriate line to /etc/postfix/master.cf  and it all seems to be  
working ok. But it doesn't seem to be very accurate in the default  
configuration - I have a mailbox with about 4,000 messages,  
approximately 98% of which are spam and it gets only about 1/3 of  
them. The statement in `perldoc Mail::SpamAssassin::Conf`that "5.0 is  
the default setting, is quite aggressive" does not seem true here.

The installation is for a small site with about 5 users who won't be  
capable of much sophistication when it comes to training spam & ham -  
it's a shame that something magical can't be done involving right- 
clicks in Outlook - so I'd just like a cheap & cheerful way to  
improve accuracy.

I'd be very happy with a 95% success rate on spam detection, but  
obviously false positives are a Bad Thing.

I read something about Vipul's Razor so emerged it before emerging  
SpamAssasin - I think this means that SA is able to take advantage of  
it - but haven't looked yet into setting it up. Is it any good? I  
read stuff about training Razor & returning feedback to the network,  
but that's really not an option here - is it possible just to use  
Razor as a "leeching" client?

Thanks in advance for any suggestions,

Stroller.

-- 
gentoo-user@g.o mailing list