spamassassin-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Maul <>
Subject Re: question about token frequency
Date Tue, 01 Feb 2005 21:55:47 GMT
Matias Lopez Bergero wrote:
> Theo Van Dinter wrote:
>> On Tue, Feb 01, 2005 at 06:20:17PM -0300, Matias Lopez Bergero wrote:
>>> That 62.56% of them was recorded once means recorded by the input of 
>>> sa-learn? or by SA autolearn?
>> Can't say, there is no difference to the database between sa-learn and
>> autolearn.
>>> I'm not very sure if the autolearn is working. How can I check that?
>> The default headers will include a "autolearn=" snippet which will let 
>> you
>> know if autolearning has occured.  If you want to manually test, you 
>> can run
>> the message through "spamassassin -D" and it'll tell you what is going 
>> on.
> ok.
> I receive a message marked as spam with an score of 19.10.
> In my SA the auto learn threshold is set to 12.0 for spam, and 
> I can't see the auto learn snippet on the message headers.
> That means that auto learn isn't working?
> I'm going to try passing the message through spamassassin with the debug 
> flag and read the output.

You should see something like:

X-Spam-Status: No, hits=-4.9 required=5.0 tests=BAYES_00 autolearn=ham

However, the score being higher than the autolearn threshold is not 
enough to determine whether or not it is autolearned.  The score has to 
be high enough, but there also has to be a certain number of points from 
scores that check headers and a certain number of points from scores 
that check body.  I dont know the exact numbers but if you look around 
on the sa site im sure you'll find them.


View raw message