Sunday, November 27, 2011

Reaction: Jigsaw: Supporting Investigative Analysis through Interactive Visualization

It is interesting to see a work that puts the analyst in charge of analysis than have an algorithmic and automated system. There must be pros and cons of each method and makes me wonder what they would be.
Jigsaw gives benefit to the user since it gives different options for them as it provides multiple perspectives (tabular connection, semantic graph, scatter plot, and text view) on a document collection. It also translates an event to other views to communicate among the different view which helps to understand the relationship better. In addition, providing a search function is very useful.
Figure 2. List view has a scalability issue where multiple columns of entities or a lot of relationship among the entities would clutter the whole view. I wonder whether "show all connections" would help analyze the information when there are numerous relationships. Although they provide "move active up" to kind of solve one of its limit, but it does not provide a general solution. Again, graph view has its limit on scalability as the number of nodes and edges grow. Filtering the information should be provided to the user. With scatterplot view, clutterness of multiple representations (diamonds) may mislead analyzing the accurate information since they appear together in the view at the same time. And yes, they mention about this limit in discussions.
It is necessary to show what the color of the diamond and red colored text mean at the side of the view. Or is the system for a particular analyst who looks at the views? As the last paragraph of 3.5 talks about user applying particular color to a report, would the user specify colors to his or her interesting document/report and then start analyzing?
It would be interesting to use this system to see different views for a E-R(Entity-Relationship) diagram.