Manual Page

Table Of Contents

  1. Introduction
    1. Platforms
    2. License
  2. Using The Extension
    1. Installation
      1. Simple Mode
      2. Advanced Mode
    2. Features
      1. Two Modes
      2. Accounts and Profiles
      3. Classifying a page
      4. Information Displayed
  3. Download
    1. For Linux

1 Introduction

Burfiks is an extension of the popular browser Firefox. It collects the links in the active web page and using preferences of the user classifies and presents them along the lines of anti-spam Bayesian filters. This tool is extremely useful to navigate quickly across pages that contains many links, as it identifies quite reliably the most interesting ones in real time.

This extension implements a statistical approach to improve the quality of the results returned by web search engines, and more generally, to sort the links of a web page according to their interest.

An inevitable consequence of the ambiguity of the natural language used to express search queries is that interesting results of searches on the web are often hidden among many unrelated matchings. use of endogenous webs (among the documents that match a query, the most relevant is the one which is pointed by the greatest number of them) as done by Google turned out to be a promising solution.

However, as far as we know, an interesting and simple approach has not been considered yet: search by Bayesian filtering. Bayesian filters have been made popular by Paul Graham to identify automatically Spam messages. They rely on an elementary theorem of probability theory known as "Bayes formula". Intuitively, by examining old messages that have been first classified by the user into "Spam" and "not-Spam", they build a probability distribution that has good properties to predict if an incoming message will be classified by the user as Spam or not.

This project applies the Bayesian approach to refine the results returned by web searches, and, more generally sorting the links found in a web page according to their interest.

1.1 Platforms

The Burfiks package is available for Linux, Mac and Windows operating systems. But currently for the development phase, we are distributing binaries only for i686 GNU/Linux.

1.2 License

Burfiks is distributed under the terms of the GPL (GNU Public License) version 2, available from

2 Using The Extension

The extension has thee parts.

2.1 Installation

The extension is being distributed in two formats.

2.1.1 Simple Mode

In this mode, all the three parts are bundled into a single xpi package. Download the xpi file from the website and drag it onto Firefox. Firefox will then ask for a confirmation. In this mode the server and the proxy server will run on the clients machine. They will be launced by the sidebar and the user need not do anything.

2.1.2 Advanced Mode

In this mode the server and proxy server binaries are distributed separately from the sidebar client. The user will have to specify the location of the server to the sidebar. This is particularly useful for groups and organizations as then not every client machine will need to run the server. To run burfiks, install the burfiks.xpi extension in Firefox, and execute one of the following in a shell:

2.2 Usage and Features

After the extension is installed, burfiks runs as a sidebar in the firefox browser. When a user visits a page, all the links in that page are collected and the corresponding pages are dowwnloaded by the proxy server (polipo) and based on the content of the page, the links are ranked in real time. To help in the ranking of the pages the user classifies the pages he visits as interesting or not.

2.2.1 Two Modes

You can select either the default simple mode or the advanced mode(discussed above) to specify how the server is launched. In the advanced mode, you will have to specify the host and port the server is running on. In the simple mode, the server is lauched and shutdown automatically.

2.2.2 Accounts and Profiles

The user has to create a account to start using the extension. The user can then create several profiles to specify the interests like personal, work etc. So different profiles give the corresponding links ranked according to that profile. The user can create and delete profiles from the sidebar. Also the account of the user is password protected so that other users cant access your usage patterns. As the server and profiles information lies on the local machine, the user has better privacy. The user can also delete his account and all the profiles and their information are simultaneiusly deleted.

2.2.3 Classifying a page

To train the bayesian classifier to the interests of the user, the user has to classify the pages whether the page interests him or not in the sidebar. Then the words from the page are used to train the bayesian classifier so that pages can be ranked.

2.2.4 Information Displayed

The sidebar displays and orders the links according to their ranks and displays the list of top 15 keywords which were considered by the bayesian while ranking the page. The user also has the option to view his profile-specific database to get an idea of the words which have been attached to him/her.

3 Download

3.1 Linux