The FilterGus Project

Overview

A simple Java Program, that filters up any kind of textual documents, provided with Profiles of the user's actual interests. It is a Information filtering tool, that compares every document against a profile previously loaded.
So you have the documents, and the profile. The latter is one but is made up adding several .gus files in a unique profile, that teach FilterGus what exactly you are looking for.
So, any session is simply made up with these steps:

build the profile loading the .gus files you need (usually one for the language, one for the document structure -these are loaded automatically- and one or more for the contens) or create a new profile from scratch showing FilterGus a document (the feedback function) you liked the most.
now feed him with the documents to be filtered, he will show a score beside each one when it's done. Select the link with the mouse to control it with the action buttons.

Don't forget to load in the .gus file about the content, otherwise all the scores will be zero! If you want you can instruct FilterGus to always load some files at start up, just open the startup.gus file with a normal text editor and add the wanted files after the others, using the same syntax. With this mechanism you can change language and/or document type. This release is shipped with standard english and HTML files only.
Remember that .gus files are additive, so you may filter up documents in different languages and/or for different formats simultaneuosly.

For which kind of documents..

Any kind of textual document; from emails to XML files and whatever. In the following we will talking almost exclusively about HTML files but keep in mind that it's just one of the uses.

An Example

This is just a dummy example.
Let's suppose that I've found an interesting article about, say XML, and I want to find more on that topic. I show this document to FilterGus, and pushing the feedback button he will create a .gus file about the contents of the document.
Then I go surfing around, but first I load that .gus file using the load button, checking if it's alright pointing with the mouse just above the window corner down to the right and and after a while all the loaded files will show up. pointing to the bottom-right corner shows the currently loaded profile

Then I surf around, try to figure out a good starting place let's say at the XML page at w3.org, and I submit this URL to FilterGus; while I'm browsing the page, he's filtering it out scoring each document against the one I previously loaded in.

Documents are visited according to different strategies, here breadth-first

After a while I found the best page about what I was looking for, and jumped there.

also the branching depth could be modified, here is set to one

By the way, I know not only poor rated links but also the broken ones or those temporarily down at that moment from my place.
And that's just one little example of the many possibilities of FilterGus..
I should warn you anyway that this is a work in progress and not all the features are perfectly implemented yet. The algorithm itself has been developed theorically just for this purpose and then implemented using the latests from the Java language. FilterGus was the first full-working Java Applet on Information Filtering, and probably still it is.

Applet and Application

FilterGus can run in both modes.
Of course the Applet one is better because you can browse as you usually do and FilterGus will filter the pages from your browser automatically, just as a kind of plug-in. The application mode has not this strong link with the browser and you need a bit of work more to specify the documents to be filtered.
Anyway be conscious that the Applet needs Java Plug-in^tm from SUN in order to work.

Applet

use the previous release (0.1)

Application

download and unzip the gus.zip file,
add to the classpath the path of the jar archive, or just type in the following line:

cd c:/mypath/gus

java -classpath c:/mypath/gus/filtergus02.jar gus/GusFrame

(mypath is just to be substitued with your actual path)

Commands

here are some of the features:

extensive usage of the tooltip facility (that little balloon that springs out when you point something with the mouse) in order to express not only help text, but also helpful data without filling up the screen with stuffs. For instance go to the left corner of the window (just on "loaded profiles") and the springing text balloon will notify you all the profile you've loaded since.
a status bar (bottom-left) will provide informations about the selected link.
at any moment you can stop or restart an already stopped link (a node in the tree)

Installation instructions

You have tho choices with the same program: running it as a standalone application or as an applet. Depending on your choice also the installing requirements change.
This software needs to be launched locally, so first you download the whole package, unzip it in your favourite directory, then it's ready to go. To use it within your browser (as an applet) visit this page before starting your searches; it will load FilterGus that will assist you filtering out the documents.

application

Requirements

JRE 1.1.6 (or 1.2) or JDK1.1.6 (+Swing, or 1.2)

applet