To open this window, click the menu at the top of the and then click Preferences.... Then click the Filters icon at the top of the window and then click the Bayesian tab at the bottom of the window.
See also How the Bayesian Filters work.
The two lists in this window show you the most significant spam and non-spam tokens, based on your personal email history. These tokens are taken from email you have actually received. Thus, each person's tokens will be unique to that individual.
You cannot edit this information but sometimes it is fun to see the words that appear most frequently in your spam and good email.
Sometimes this information can be used to troubleshoot problems. If the Bayesian filters are very inaccurate, you can look in this window to see what the significant tokens are. If you see anything strange in this window—like "viagra" in your non-spam tokens (and you do not work in pharmaceutical sales)—it may indicate that you have trained Spamfire incorrectly.
You can check Treat spam scores of 1 and 99 as "Definite" to cause email messages with extreme Bayesian scores to appear in the "Definite" mailboxes in the mailboxes window. Normally, only messages that get identified by a Sender, Rule or Remote filter will appear in the "Definite" mailboxes. You can turn this setting on after you have been using Spamfire for a while and it is doing a good job of identifying good email and spam. It will cause more mail to appear in the "Definite" mailboxes. Then you will find it easier to concentrate on the messages stored in the "Probable" mailboxes. When you turn this checkbox on, settings you make in the Delivery Preferences for "Definite Good" and "Definite Spam" mail will also work on messages with extreme Bayesian scores.
See the icon legend for a definition of "Definite Spam" and the other email types in Spamfire.
If you think you have mistrained Spamfire by accidentally marking spam messages as good or vice versa you can erase all training and start over again by clicking Reset the Bayesian Corpus...
Spamfire is able to recover from occasional mistakes and oversights. As time goes on, small errors in training will have less effect on Spamfire's accuracy. Only use this option if you think you have made really major errors in training Spamfire.