sa-learn: lower the 200 thresold

[Mário Gamito <mg...@telbit.pt>]

Hi,

How can change the number of messages needed for sa-learn from 200 to a 
lower value ?

My boss (grunf... it had to be him) is getting a lot of HAM.

Any help would be appreciated.

Warm Regards
-- 
:wq! Mário Gamito

Re: sa-learn: lower the 200 thresold

[Theo Van Dinter <fe...@apache.org>]

On Fri, Apr 13, 2007 at 03:01:04PM +0100, Mário Gamito wrote:
> How can change the number of messages needed for sa-learn from 200 to a 
> lower value ?

You can edit the code, but why would you want to?

> My boss (grunf... it had to be him) is getting a lot of HAM.

Ok, I get a lot of ham too.  So what?

-- 
Randomly Selected Tagline:
"Yeah, everybody trusts the cops since they went after Rodney like a
 pinata."                  - Dennis Miller

Re: sa-learn: lower the 200 thresold

[mouss <mo...@netoyen.net>]

Mário Gamito wrote:
> Hi,
>
> How can change the number of messages needed for sa-learn from 200 to 
> a lower value ?
>
> My boss (grunf... it had to be him) is getting a lot of HAM.

if "it" doesn't come to you, go to it! do it the other way. find 200 ham 
and 200 spam messages and sa-learn them! even if these are different 
than your boss email, it's still better than changing the code...

>
> Any help would be appreciated.
>
> Warm Regards

Re: sa-learn: lower the 200 thresold

[Matt Kettler <mk...@verizon.net>]

Henrik Krohns wrote:
> On Fri, Apr 13, 2007 at 10:32:56AM -0400, Matt Kettler wrote:
>   
>> Mário Gamito wrote:
>>     
>>> Hi,
>>>
>>> How can change the number of messages needed for sa-learn from 200 to
>>> a lower value ?
>>>
>>> My boss (grunf... it had to be him) is getting a lot of HAM.
>>>
>>> Any help would be appreciated. 
>>>       
>> Edit the code.
>>
>> That said, you may get unreliable and erratic bayes results from SA if
>> you do this. There's a reason the minimum is 200 and there's no
>> adjustment. You can't make a reasonable statistical analysis without
>> enough data to base it on.
>>
>> If your boss is having mis-tagging problems, I'd look elsewhere. This
>> will make the tagging problems worse, not better.
>>     
>
> Umm, it would help learning how to do basic configuration or atleast
> find the manual.
>   

Gah! Alas, I am undone :)

Re: sa-learn: lower the 200 thresold

[Henrik Krohns <he...@hege.li>]

On Fri, Apr 13, 2007 at 10:32:56AM -0400, Matt Kettler wrote:
> Mário Gamito wrote:
> > Hi,
> >
> > How can change the number of messages needed for sa-learn from 200 to
> > a lower value ?
> >
> > My boss (grunf... it had to be him) is getting a lot of HAM.
> >
> > Any help would be appreciated. 
> 
> Edit the code.
> 
> That said, you may get unreliable and erratic bayes results from SA if
> you do this. There's a reason the minimum is 200 and there's no
> adjustment. You can't make a reasonable statistical analysis without
> enough data to base it on.
> 
> If your boss is having mis-tagging problems, I'd look elsewhere. This
> will make the tagging problems worse, not better.

Umm, it would help learning how to do basic configuration or atleast
find the manual.

http://spamassassin.apache.org/full/3.1.x/doc/Mail_SpamAssassin_Conf.html

bayes_min_ham_num (Default: 200)
bayes_min_spam_num (Default: 200)
    To be accurate, the Bayes system does not activate until a certain
    number of ham (non-spam) and spam have been learned. The default is 200
    of each ham and spam, but you can tune these up or down with these two
    settings.

That said, no comments on whether it makes any sense to lower it..

Cheers,
Henrik

Re: sa-learn: lower the 200 thresold

[Matt Kettler <mk...@verizon.net>]

Mário Gamito wrote:
> Hi,
>
> How can change the number of messages needed for sa-learn from 200 to
> a lower value ?
>
> My boss (grunf... it had to be him) is getting a lot of HAM.
>
> Any help would be appreciated. 

Edit the code.

That said, you may get unreliable and erratic bayes results from SA if
you do this. There's a reason the minimum is 200 and there's no
adjustment. You can't make a reasonable statistical analysis without
enough data to base it on.

If your boss is having mis-tagging problems, I'd look elsewhere. This
will make the tagging problems worse, not better.