Researchers Maneesh Agrawala and Fernanda Viegas walked through some amazing technologies that they've been cooking up in their labs for data visualization. As a data junkie, I was totally mesmerized. The Wikipedia analysis was an particularly fascinating visual interpretation of living content. (Disclosure note: Omidyar Network is an investor in Wikia, a related entity.)
Reading the Wikipedia genome
IBM's Collaborative User Experience Research Group has developed an insightful system for showing how content evolves. They can interpret the ebb and flow of content over iterations using a variety of contexts, such as time, version, or author. Wikipedia serves as a perfect complex subject for analysis.
In Viegas' demo, it
was especially dramatic to look at the visualized iteration of more
controversial pages, such as abortion or god. You could see the flow and
rhythm of changes such as revision wars, deletions, or reworking.
Some other tricks that stood out:
Seeing someone's history. See all of the subjects that someone has written or talked on, and judge whether or not he/she is a credible editor of the content that you have written. This works like a tag cloud, and the revision spectrum can be sorted by title, content, namespace, or history.
Interpreting semantic meaning. Categories make more sense when clustered as vs. listed alphabetically.
As it turns out, Wikipedia has since disabled the export page that was needed to perform this analysis. You can still can perform test analyses using the archived information, however, and the team is working with Wikipedia to figure out a way to perform visualization without creating the tera-plus downloads that kill the Wikipedia servers. Download the history flow application here.
The most interesting thing that the researchers discovered through this work was that visualization provided a means for people who had never touched Wikipedia before understand the dynamics of the community. At FOO, we also talked about how this technology opens up new possibilities in identity systems. Is your online history as personally identifiable as a fingerprint or a brain scan?
Maps optimized for humans
Agrawala's maps visualization grew out of frustration with driving directions on services such as MapQuest. The maps themselves are not usable as driving directions, so people typically rely upon the written directions rather than the map. Comparing a Mapquest map to a hand-drawn map highlights several key elements of usability in the easy-to-follow, hand-drawn map:
- Exaggerated road lengths
- Regularized turning angles
- Simplified road shapes
These subtle characteristics emphasize the important information in the map. The Mappoint system, called LineDrive, automates the characteristics of a hand-drawn map. I was most intrigued by the system's trick of emphasizing the size of short roads in the map in order to optimize usability. For example, the path from the highway exit to your destination is not drawn at the same scale as the highway itself. Those surface roads are larger, so that you can see the complicated turns more easily.
While Agrawala's team is focused on visualization, their photomontage system also provides more ways for people to use photographs. Photomontage allows you to select a set of images, then draw some coarse strokes that tell the system which areas you want from each picture. The system then computes which region it will take from each photo in order to create the final image, and how it will reassemble the final image.
Some current applications:
- Correction of bad lighting
- Continuous depth of field in digital photography
- Better specimen identification - microscopes have the same problem that cameras do, and the photomontage system allows scientists to see the full specimen clearly
- Removal of moving objects from a landscape - like TouristRemover, you can remove all of the people from a set of photos that in aggregate shows all portions of the landscape
There was some controversy around this technique, which folks thought would be used to create fake memories. For the everyman user, though, it's more likely to result in a holiday photo in which everyone is smiling and facing the camera - which is what your mind's eye remembers anyway.