Information Visualization

05/28/02 from aboutAI.net

The good old days of command line interfaces are passing quickly. Automated information visualization techniques are needed to fully exploit the benefits of knowledge management tools. Learn what is available today and how it can be put to maximum use.

We had several queries related to the information visualization techniques recently posted to our forum. Naturally, knowledge management and other similar AI-related fields could greatly benefit from such "smart" visualization tools. To quote Ben Shneiderman ("Dynamic queries, starfield displays, and the path to Spotfire"), ... the good old days of command line interfaces and submitting queries to databases are passing quickly. Knowledge management investigates the various issues which organizations face to collect, store, retrieve, review, and use knowledge critical to organizational performance. However, the explosive growth of the Web and offline information repositories resulted in a need for new strategies for efficient management of vast quantities of available information. Automated classification and clustering techniques are needed to fully exploit the benefits of both directories and search engines. Dick Stenmark illustrates these issues in the abstract of his excellent paper "To Search is Great, to Find is Greater: a Study of Visualization Tools for the Web":

Visualization of the AI part of the cyberspace, as mapped by the Open Directory and run through the WebMap's engine

...Meanwhile, the user is left with a search interface consisting of one input field (in which they have no idea of what to type) and one push button. Hardly hi-tech. Most users type in a single keyword, click on the button, and hope for the best. It seldom occurs. Desperately needed are better designed user interfaces that utilize natural language and agent software to guide the user through the iterative process of formulating the query. Add to that state-of-the-art visualization techniques and relevance feedback to help with navigation and selection in the too often crowded result set, and we may indeed end up with something truly useful.

If we disregard the related issues of natural language understanding, query expansion and domain pruning, we'll quickly find that the visualization techniques could offer the key ingredient needed to handle the problems described above. Cone trees, hyperbolic trees and bullseye view are all techniques well suited for displaying hierarchies, such as organization charts and link structures of Web servers. Steinmark's paper also describes several novel solutions, including perspective walls, PARC's Scatter/Gather, detail view tools, and other hybrid solutions.

Antarcti.ca's look at the AI world

Several companies are offering commercialized implementations of these algorithms. Amongst the most popular are Antarcti.ca's Visual Net software, a next-generation knowledge management tool; Thinkmap, a tool for displaying complex information using an animated multidimensional display designed for user interaction using Java; TheBrain, an easy-to-use system for organizing and sharing information; SemioMap, a tool for discovering relevant information from seemingly impenetrable mountains of text; Inxight's Star Tree and Table Lens, products for navigating and visualizing Web sites and other hierarchical information collections and datasets.

As for the technical background, several solutions rely on the information visualization via hyperbolic geometry. Hyperbolic space has "more room" than Euclidean space, allowing more information to be seen amid less clutter, while motion by hyperbolic isometries provides for mathematically elegant navigation. However, Inxight holds patent rights in this area, but several good open-source solutions like the Hyperbolic Tree Java Library still enables you to experiment with this approach.

On a side note, topic maps are a new ISO standard for describing knowledge structures and associating them with information resources. As such, they are often described as the "GPS of the information universe", providing new means of navigation through large and interconnected information spaces. The topic map market is quickly maturing, with a number of vendors providing industry strength topic map tools. The leading providers of topic map solutions are Ontopia (Ontopia Knowledge Suite), Infoloom (Topic Map Loom technology), Empolis (empolis k42) and Mondeca (Mondeca Organizer, Mondeca Navigator, Mondeca Viewer).

My PersonalBrain

Information visualization tools are traditionally implemented as a set of pre-defined visual displays. An alternative, interactive approach taken in the DOODLE project supports visualizations specified by the user with a visual constraint-based language. Another rather exotic approach for visualizing information spaces is provided by the WEBSOM, self-organizing map for Internet exploration. It is a method for organizing miscellaneous text documents onto meaningful maps for exploration and search. WEBSOM is based on the SOM (Self-Organizing Map) algorithm that automatically organizes the documents onto a two-dimensional grid so that related documents appear close to each other. Unfortunately, there are no downloadable software here, but I'm currently working on something similar using the JavaSOM package. The visualization of results (the trained map) is implemented by using Scalable Vector Graphics (SVG), which is an XML application language.

There are many great online resources that will guide you to the world of information visualization, and I simply cannot describe all of them in a single article. A quick review of the archived links related to this area (and stored in my PersonalBrain ;)) reveals a Chapter 10 of the book Modern Information Retrieval by Marti Hearst, dedicated to User Interfaces and Visualization. Another excellent source of information is maintained by Katy Borner, assistant professor at Indiana University: don't miss her research page and Information Visualization Software Repository. Mappa.Mundi Magazine is another indespensible resource that investigates how researchers help define our perception of cybergeography and virtual spaces. Other recommended sites full of detailed information are the Information Visualization hub at Open Directory, InfoVis.net, An Atlas of Cyberspaces, Visualization & Clustering Tools at SearchTools.com and OLIVE, On-line Library of Information Visualization Environments, and an extensive list of Information Visualization Resources.