Spam Blocker   anti - spam - tools .com
  Spam Filter Filter Spam without receiving it!

Mail Box Dispatcher 2 - Self-learning Spam Detector


Self-learning Spam Detector

Spam Detector is a self-learning artificial intelligence that calculates spam probability for all incoming messages. To calculate spam probability ratings Spam Detector uses a special algorithm (optimized Bayes probability). The algorithm applies to the results of statistical analysis of YOUR email messages and YOUR meaning of wanted and unwanted correspondence, that empower to determine spam with very high accuracy.

When, How and What does it learn?

Spam Detector observes incoming correspondence and extracts textual words, useful numbers and other information (such as presence of long words, absurd alpha-numerical combinations, domains and much other) using advanced tokenization mechanism.
When you mark messages for deletion and click Process, Self-learning Spam Detector learns the deleted messages as unwanted and saves all the received info of unwanted messages into the 'spam' dictionary. Info of other messages goes into the 'good' dictionary.
If you delete a message that have been learned as 'good', the intelligent self-learning Spam Detector unlearns it and then learns it as 'spam'.
When you exit from the program, Spam Detector learns all messages that left in your mail boxes as 'good'.
You can see word and dictionary statistics via Spam Detector tab of the Options menu (Tools > Options).

Learning takes time

As you know learning takes time. Spam Detector always learns. It even doesn't show you spam probability until it learns 20 'good' and 20 'spam' messages to not confuse you with its initial naive forecasts. But with every new learned message Spam Detector improves its detection abilities and will tell you more truthful spam probability rating for incoming email messages.

Self-learning Spam Detector is one of an excellent Mail Box Dispatcher features that help you to filter spam without downloading it.


Copyright © 2003-2006 Alex Kaul . All rights reserved.