Koozali.org: home of the SME Server

Legacy Forums => General Discussion (Legacy) => Topic started by: widman on April 25, 2005, 04:55:19 PM

Title: Many Thanks and One Question
Post by: widman on April 25, 2005, 04:55:19 PM

First I would like to thank the Contribs community for a great product.  I have no experience with Linux and managed to set up an email server, complete with the Spam filter and Antivirus.  It took a little digging, but I found almost all my answers in the documentation or the forums.

I do have one question concerning SpamAssassin.  I have 3 directories accumulating Ham, Spam and missed spam.  I'd like to train the Bayesian filter to improve its accuracy, but I am unsure about which mail to have SpamAssassin learn.  The Ham and Spam folders contain mail that SpamAssassin has already filtered correctly.  Is it useful to run sa-learn on correctly filtered email or should I just run sa-learn on the missed spam?

thanks,
Pete

Title: Many Thanks and One Question
Post by: Jon_Reynolds on April 25, 2005, 11:06:37 PM

Don't quote me on this but I don't think you will have a problem. There was an issue in a previous version of SA that would cause problems if the headers were already marked up by SA but that has been resolved.

I would go ahead and do it but wait till you have about a 1000(thousand) of each to run it so you have a good corpus.

Hope that helps,

Jon