[lug] Spam question

[Sean Reifschneider]

On Tue, Mar 13, 2007 at 09:58:44PM -0700, karl horlen wrote:
>How could this poison a baysian filter (even in
>theory)?

If you have nonsense text and someone tells their bayesian filter that it's
spam, it makes the bayesian filter worse.  Because it is not the text that
is pushing the product or service, it doesn't help the Bayesian filter
block future mailings.

If you were instead OCRing the GIF attachments that I usually see
accompanying these messages (or did, before I started blocking all messages
with GIF attachments), and using that for Bayesian training and checking,
you'd get much better success because it includes things like:

   Stock ticker symbol: XYZY
   Target selling price: $69

Sean
-- 
 Dear Santa, all I want for Christmas is your list of girls who were naughty.
Sean Reifschneider, Member of Technical Staff <jafo at tummy.com>
tummy.com, ltd. - Linux Consulting since 1995: Ask me about High Availability
      Back off man. I'm a scientist.   http://HackingSociety.org/