On Sat, 2011-11-26 at 19:59 -0500, Ross Vandegrift wrote:
On Sat, 2011-11-26 at 19:10 -0430, Patrick O'Callaghan wrote:However your problem seems to be the reverse, i.e. you're getting spam in your inbox. In that case, it's a strong indicator that the filter has not learned enough to distinguish between ham and spam.Is there any way to see the output or results of the spam filter run on a message? Evo has almost never marked a message of mine as Junk, but I don't see anyway to troubleshoot it.The way to deal with that is to train it. The exact method varies according to the filter, but with Bogofilter (the one I use), the easiest way to to save a bunch of spam messages (the more the better but around 100 is a good number) in an mbox file and run "bogofilter -M -s < file". Do the same with a bunch of non-spam messages (using '-n' instead of '-s') and you'll have an initial corpus that Bogofilter can use.Wait - are you saying that "Mark as Junk" doesn't do that for me?
Following up to myself: I've done some experiments today, and at least for the spamassassin plugin (I don't use bogofilter), the "Mark as Junk" and "Mark as Not Junk" do indeed train the bayesian filter. So if you're marking enough messages, there should be no additional need to train the filter. Still unclear on how to troubleshoot spam filtering in Evolution - I figured this out by looking at top and watching the SA db magic change. Ross -- Ross Vandegrift <ross kallisti us>
Attachment:
signature.asc
Description: This is a digitally signed message part