How does the autolearn feature work?

["M. Possamai" <po...@xs4all.nl>]

Hi,

I was wondering..
How does this autolearn thingy work?
What will happen when I turn it on?

I mean, how will it know what's spam and what's not spam?

Re: How does the autolearn feature work?

["M. Possamai" <po...@xs4all.nl>]

At 22:04 29-11-2004, you wrote:
>At 02:44 PM 11/29/2004, M. Possamai wrote:
>>How does this autolearn thingy work?
>>What will happen when I turn it on?
>>
>>I mean, how will it know what's spam and what's not spam?
>
>It calclulates what score it would have gotten in a non-bayes scenario. 
>There are ajustable learn thresholds. If it's under the ham threshold, 
>it's learned as nonspam. If it's over the spam threshold it's learned as 
>spam. Anything inbetween is skipped.
>
>(note: there are some additional conditions on learning as spam, such as 
>learning as spam requires 3.0 of header rules and 3.0 of body rules, in 
>addition to the overall pre-bayes score being over the threshold. They all 
>amount to basic paranoia to prevent learning mis-classified mail as spam.)

Awsome explaination.
Thanx

Re: How does the autolearn feature work?

[Matt Kettler <mk...@evi-inc.com>]

At 02:44 PM 11/29/2004, M. Possamai wrote:
>How does this autolearn thingy work?
>What will happen when I turn it on?
>
>I mean, how will it know what's spam and what's not spam?

It calclulates what score it would have gotten in a non-bayes scenario. 
There are ajustable learn thresholds. If it's under the ham threshold, it's 
learned as nonspam. If it's over the spam threshold it's learned as spam. 
Anything inbetween is skipped.

(note: there are some additional conditions on learning as spam, such as 
learning as spam requires 3.0 of header rules and 3.0 of body rules, in 
addition to the overall pre-bayes score being over the threshold. They all 
amount to basic paranoia to prevent learning mis-classified mail as spam.)