Okay installed the 2.0.1.302 version and all seems to be going well.  I know the new bayesian filter does not kick in untill we hit 5000 in the corpus.  We are at about 3,500 as I write this, with 560 being marked as spam.  Now once it does kick in and starts putting the messages it thinks is spam how do I update the corpus to tell it about false positives and missed spams?  If I forward a message from Quarantine tab with the Deliver Selected button does that update the corpus (I have a feeling it does not) then how do I go about changing the corpus.  I know you stated about a program that will let us do that, is that what I will have to wait for?

As always thanks for a great program and as I have seen in the forum fast response to feature updates (ie: attachment filtering which I have already put to good use).

Hillard

Hillard,

SpamFilter actually already sees that emails have been marked as false positives when messages are "Deliver Selected" from the GUI or from the web interface. When these events occur, the corpus is updated to reverse-mark those token and SpamFilter will apply extra "weight" on them in the corpus to help reduce the false positive on similar emails in the future.

Please also note that you do not have to wait for the 5000 to kick in before doing this. The reversal process will work at anytime, even during the learning progress.

Roberto F.
LogSat Software

Thanks, that is great and answers that question, this only works for the bayesian caught messages not the keyword filter messages?  How do the bayesian caught messages ident themselves in the web interface and the gui?  Since all I see are marked by being caught by my keyword filters, which leads me to the question below. 

Also, in the forum there is discussion about not having anything in the keyword filter blacklist?  I left my entries there, should I take them out?  Also when the spam gets through the bayesian filter how to I submit it so it does not come through again, esp if I have to take out all the keyword filters in order to get the bayesian to block it.

As always thanks.

Yes, that applies to all filters. In particular, the messages caught by the Bayesian filter will have a RejectID value of 14, and in the web GUI the description will be "Statistical filter match".

When you force delivery of false positive emails caught by the filters, SpamFilter ISP will revert the score assigned to that message, and will also furthemore assign "extra points" to the various words to help tagging it as a "good" message next time similar emails come in.

Roberto F.
LogSat Software