Information Visualization for Knowledge Discovery

March 10th, 2009 by Sabrina
Treemap developed by Jean-Daniel FeketeTreemap developed by Jean-Daniel Fekete

Ben Schneiderman from the University of Maryland, gave a fascinating talk in Cambridge on 5th March 2009 about the topic of “Information Visualization for Knowledge Discovery.”

Ben has authored many books and papers on human computer interaction, and was the founder of the Human Computer Interaction Laboratory at the University of Maryland. His keen interest is the field of information visualisation.

During his talk, Ben pointed out that in contrast to scientific visualization information visualization is a relatively young field as information visualization conferences have only been going for about 15 years. He added that, the challenge with information visualization is that the information keeps changing over time.

Ben presented his conceptual break down of information visualization tasks: “Overview -> zoom and filter -> details on demand.” What he means is that one should provide an overview first, showing all the information, for example complex graphs, diagrams and maps. This allows the user to orientate themselves and get the big picture. Then allow the user to zoom into more detail and filter out any unwanted information. Finally, allow the user to select an item and get more detail about it when required.

The most enlightening point that Ben made during his talk was that: “Information visualization gives you answers to questions you didn’t even know.” He went on to argue that “there should be a move from opportunistic discovery to a more systematic discovery of knowledge.”

Ben illustrated his argument with a number of demonstrations and screenshots of projects that he and his students have developed over the years. Each guides knowledge discovery thought the visualization of different patterns in the data. Ben emphasised that for information visualization “the interest is not in a particular value, but an overall view and patterns in the data”. Yet, he also emphazised the importance “trying to see the violations in the data that are contrary to your expectations”.

Ben demonstrated his famous treemap that has now been modified by many commercial companies. It has been modified for example  to visually show the constantly changing landscape of the google news aggregator, and even the New York Times has used it to show changes in truck and car sales.

Other tools which Ben showed were the ShapeSearcher which finds spikes in the data, Scattergrams which provide the opportunity of hunting for stuff, the alignment tool which can filter by event and show what happened before and after this event, and another tool which identifes gaps in the data.

The most intriguing example of finding patterns in data was Ben’s demonstration of the SocialAction tool, which uncovers hidden structures in social networks over time. The visualization presented the correlation between US senators voting the same way. It showed a strong that democrats and republicans vote the same way. Only four republicans sometimes voted similar to the democrats. Yet the most surprising finding through this visualization was that the correlation for democrats voting the same way was far stronger then republican voting the same way.

"The social network of the U.S. Senators voting patterns in 2007, after Democrats took control. Republicans are colored red, Democrats blue and Independents maroon. Here, the partisanship of the parties appeared automatically (180 vote threshold)." (by Ben Schneiderman)

"The social network of the U.S. Senators voting patterns. Here, the threshold is raised to 290 votes. The Democrats' relationships are much more intact than the Republicans. Details-on-demand are provided for Senator Whitehouse, the senator with the highest degree at this threshold." (by Ben Schneiderman)

Ben concluded his talk with three key points for information visualization to guide knowledge discovery:
1.    Rank-by-Feature Framework, i.e. rank by what people want to know
2.    Decomposition of complex problems into multiple simpler problems
3.    Ranking guides discovery. It is important to provide systematic
       approaches for discovery.

Challenges of visual literacy

A theme that kept popping up in the talk and particular in the questions afterwards, was the challenge of visual literacy. Words can help to clarify matters of information visualization, but Ben explained that textual information is only good for simple queries (such as a rank list in Google search results). Visual tools on the other hand are better for complex queries.

For anyone who is interested in finding out more about the challenges of visual literacy, Ben recommended the work of Colin Ware, a perceptional psychologist, who looks at the challenges of understanding visual information.

Interesting reads about information visualisation:

Bederson, B. and Shneiderman, B. (2003) The Craft of Information Visualization: Readings and Reflections, Morgan Kaufmann Publ., San Francisco, CA. Amazon UK, Amazon US

Card, S., Mackinlay, J., and Shneiderman, B. (1999) Readings in Information Visualization: Using Vision to Think, Morgan Kaufmann Publ., San Francisco, CA. Amazon UK, Amazon US

Tufte, Edward (1983) The Visual Display of Quantitative Information, Graphics Press, Cheshire, CT. Amazon UK, Amazon US

Tufte, Edward (1990) Envisioning Information, Graphics Press, Cheshire, CT. Amazon UK, Amazon US

Tufte, Edward (1997) Visual Explanations: Images and Quantities, Evidence and Narrative, Graphics Press, Cheshire, CT. Amazon UK, Amazon US

Ware, Colin (2004) Information Visualization, Second Edition: Perception for Design (Interactive Technologies), Morgan Kaufmann Publ., San Francisco, CA. Amazon UK, Amazon US

Ware, Colin (2008) Visual Thinking for Design, Morgan Kaufman, Burlington, MA. Amazon UK, Amazon US

3 Responses to “Information Visualization for Knowledge Discovery”

  1. Sunny says:

    Sabrina, Thanks for this post.

    I think Ben Schneiderman’s work is amazing. I believe that visualization is a much better way to knowledge discovery as one can see linkage between relevant topics and even the context as well which is very important for anyone to make better sense of the information that they are reading. I enjoyed this alot.