Learning From Local User's Email

• [NOTE:]This can only be done on an email server that stores the local user's email. THIS CANNOT BE DONE ON AN EMAIL GATEWAY.

The second way in which spam and ham can be learned is by local user contributions. The requirements are that the users have IMAP accounts on the server and that they create two top level folders (folders that are in the same folder hierarchy as the INBOX folder). These folders MUST BE named "SPAM" and "HAM" (capital letters are NECESSARY). Once this is done users can copy or move their false negatives into the "SPAM" folder and their false positives into the "HAM" folder.

• [NOTE:]CHOOSE THESE USERS WITH CARE! Trusting careless or malicious users can poison the Bayes database. If the user contributes ham as spam or vice-versa, the database will operate on erroneous data and will result in operation that ranges anywhere from poor performance to reversal of email classification.

Once a user has deposited spam and ham into the appropriate folders the admin can learn these folders into the Bayes database in the Learn Users Spam/Ham Folders section of the web page. All users that have created SPAM and HAM folders can be viewed in the Select User-Name pull down menu.

Choose a user and click on View User Folders. The admin can view the email and decide whether to learn it or not. If there are any doubts as to whether the email is suitable for the database then no learning is accomplished by backing out this web page or by unselecting the appropriate check-box of Learn SPAM mailbox or Learn HAM mailbox. This is useful if the admin feels that one of the folders is acceptable but not the other. For example, if the admin approves of the SPAM folder but not the user's HAM folder she can unselect the checkbox next to Learn HAM mailbox and proceed with the learning. In this way only the SPAM folder will be learned. Only entire folders can be learned. The admin cannot selectively choose email within a folder. Once again the results of the learning can be verified by viewing the spam, ham and token counts in the statistics section of the Bayesian Learning Center web page.

If an admin feels that she has made a mistake in learning a user's email she can relearn or forget the previously learned email by choosing the user, checking the View last learned mailboxes check-box and clicking on View User Folders. A list of the most previously learned mail will be available. To relearn the email the admin can check the Re-Learn Messages checkbox. Additionally, in the Confirm Selections section of the web page the admin has to select a check-box that represents the OPPOSITE type of the folder being relearned. Another way to think of it is to check the type of classification that will be the END RESULT of the learning. For example if the user's HAM folder was previously learned as spam by mistake the admin should see this mail now represented as spam. She would then check Re-Learn Messages and Learn HAM mailbox and click on Proceed With Learning to relearn the email as ham.

To forget email check the Forget Messages checkbox and in the Confirm Selections section check the appropriate checkbox that represents the folder that the admin would like to have removed entirely from the database.