Doing this right?

[Tom Anderson]

----- Original Message ----- 
From: "JoeHill" <joehill at sympatico.ca>
>> Perhaps this person uses a mail client, proxy, etc., which inserts 
>> X-headers
>> commonly found in your spams.  Use:
>>
>> bogofilter -vvv < falsepos
>>
>> to find out what tokens are contributing to its spamicity.
>
> I looked at that, quite a lot of these:
>
> "head:MIME-Version"                7120  0.547905  0.911718  0.624626 -
>
> "head:Content-Type"                9066  0.775758  0.968561  0.555266 -
>
> I've taken a look at other messages, and it seems Bogofilter doesn't like 
> mail
> from Outlook Express, which is not at all surprising ;-)

The solution is to just keep training.  Your ups and downs should quickly 
level out.  I like to use recursive training... first train with the false 
positive, then check to see the new score -- if the new score is not in the 
ham range, train again.  Repeat.

I do recursive training (for both ham and spam) automatically with this 
program:
http://orderamidchaos.com/bogofilter/bfproxy

I also clean up my headers with this one:
http://orderamidchaos.com/bogofilter/spamitarium

Tom