The Learning Filter

Spamihilator introduces a new generation of Spam Filters: The Learning Filter!

While the Word Filter only recognizes common spam words, the Learning Filter also recognizes common non-spam words which indicate regular mails that should not be deleted.

This filter scans every mail for known words. Every word has a certain probability. With the rules defined by Thomas Bayes (an 18th century English mathematician), Spamihilator calculates whether a mail is spam or not. This filtering method is also called the Bayesian Method or Bayesian Filter.

The result of the scan can indicate a mail as spam or non-spam, or neither. If the last result is the case, the Word Filter will process the mail as well.

Training for the filter

In the beginning, the filter cannot differentiate between spam and non-spam mails. You have to teach Spamihilator in the Training Area.

When “training”, the filter will build a list of words that appear in spam mails and another list of words that appear in non-spam mails. The filter's efficiency will increase with the amount of words added to the lists when you use the Training Area.

This method ensures that the filter will be adjusted to work with your daily mails, while the Word Filter is designed to work with standard spam mails. This will increase the recognition rate and minimize the possibility of “false positives” (non-spam mails that accidentally get into the Recycle Bin).

Tests have proven that a recognition rate of more than 98% can be achieved when using the Training Area as often as possible!

 
en/learningfilter.txt · Last modified: 2008/02/26 13:16 by michel