Anti Spam Landscape

[Andy Bradford]

Thus said Peter McNabb on Mon, 23 Jul 2007 05:51:44 MDT:

> However,  other  users who  do  the  training  are only  getting  ~70%
> accuracy.

I  have found  that  those with  low  accuracy are  not  as diligent  in
retraining message that are classified incorrectly. A bayesian filter is
only as good as the trainer. :-)

> There is  a way to set  up a group  account that everyone uses,  but I
> haven't yet tried that.

Group filtering,  at least  with a  statistical approach  like bayesian 
filtering, will not be accurate at all. Most people do not get the same 
types  of  spam  and some  words  that  are  spamish  for one  may  not 
necessarily be so for another.                                          

Andy
-- 
[-----------[system uptime]--------------------------------------------]
 10:20pm  up  3:40,  1 user,  load average: 1.25, 1.18, 1.18