Filter for marking inappropriate content on web pages

Abstract

Modern web sites often contain visitors' comments which are inappropriate, offensive or even violent. It is however possible to avoid having to read such text with appropriate tools. We developed one based Bayesan techniques used in spam filters. The goal was to teach the web browser to alert the user about any inappropriate content on the web page. To make the tool user friendly, we used a number of different technologies packed in a Firefox add-on. This work describes the used technologies and their relations. We start with a description of SpamBayes, the most efficient word-based spam filter at the moment, o which we based our work. In the second part we describe the extension itself, with all details and tricks to make it more efficient. The resulting tool confirms the usefulness of the approach to filtering inappropriate messages. For this reason, we will further develop the presented add-on, and we hope that it will atract new users and, possibly, developers.