Re: problems with spamprobe

Brian Burton wrote:

> The hash format uses mmap() which does not seem to update the
> timestamp of the file on Linux.  Not sure why.  To be certain that
> nothing changed in the file you would need to md5sum it before and
> after running the training command.
> 
>> I've collected a bunch of spam and a bunch of non-spam and tried to
>> teach spamprobe by doing:
>>
>> spamprobe good /my/non-spam/maildir/
>> spamprobe spam /my/spam/maildir/
> 
> Is maildir in your example a directory name?  Is it really a maildir
> directory (i.e. with .cur and .new subdirectories)?

yes, it is a directory name and yes it is really a maildir directory.

> Try running with -v and see if spamprobe prints information about the
> emails it is processing.  If it doesn't then it's not finding any
> email to learn from.

It did give some info.


> In any case it's generally better to use auto-train or auto-learn for
> initial training rather than separate runs of spam and good.
> 
> spamprobe auto-train -v GOOD /non-spam/maildir/ SPAM /spam/maildir/
> 
> All the best,
> ++Brian

Just for kicks I tried using the BDB database instead of the hash one 
and now spamprobe seems to be learning and is able to correctly classify 
some mail.  As near as I can tell, I'm not doing anything differently 
other than having picked this different configure option.

Thanks
-Dan


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/