Sunday, November 27, 2011

Reaction: Jigsaw: Supporting Investigative Analysis through Interactive Visualization.

This is one of the few papers I have read which actually tackle the issue of growing data sets. New correlations will come into existence with larger data sets which may not be quite evident with the smaller ones - which were initially used as a basis to build a tool for visualization. Ensuring more control on the analyst's side certainly helps create new correlations and visualize new data sets but at the same time the scope for committing an error considerably increases. This comes with an added cost of time and the labor charges incurred. I am impressed with the authors aim to reduce this error margin by using a new tool called the Jigsaw. It is however interesting to note how many different types of correlations between data sets can Jigsaw actually unearth?

Another interesting point which I have noted about Jigsaw is the number of different views it supports to visualize the data. I did not find the List view to be of much use when there are many connections between entities and the page displays only a limited set at a time. By the time the user actually gets to a matching row in B from A, the original row in A can be hidden or the line connecting A and B goes absurd. The graph view is my favorite out of all the views. In particular, I liked its incremental approach of presenting a graph starting with a smaller network and increasing the level of detail in the graph depending on the users selection.

I found the scatter plot view to be a little confusing, given the coloring scheme and the spatial location of all the diamonds (reports) presented, however the density of similarity between two texts is best viewed in a scatter plot. I am also not very impressed with the text view, given that the paper was published in 2007, there are already existing tools which do a much better job in denoting similar entities with different colors for different range of frequencies.

I liked the scenario at the end of the paper especially because it describes how all the different views are used together. Another impressive aspect about Jigsaw is how the user actions from one view are propagated to another view thus saving considerable time and effort of the analyst. Overall, I felt the tool is in its very early stages of development and it would be interesting to see how the entities actually relate to each other when dealing with big data sets. I liked the paper because of the authors simplicity in presenting the material and the functioning of the tool is presented with screen shots of the outputs based on a sample scenario which definitely helped me understand the material much faster.