Tuesday, November 15, 2011

Reaction: Balancing systematic and flexible exploration of social networks

It is impressive to see most of the previous researches and efforts that went on visualizing graphs. It reflects how much people need graph visualization and how much they want to improve the layout.
The confusing part with color coding in Figure 1 is that red color starts from bright red to dark red while green color starts from dark green to bright green. The rank decreases as the list goes down but the brightness of each color confuses me. Later on, the paper describes, "each ranking assigned a corresponding color ranging from green to black to red", but still I think brightness changing in opposite way for red and green can mislead the interpretation of ranking. It is interesting and convinces that the system is useful as it shows "Muslim Militants" in bright red with highest ranking, which seems to be a hub for India, Philippines, Indonesia, Algeria, and Egypt.
The functionality of the tool explained in the paper shows that zoom-in/out and filtering are crucial to be equipped in a visualization tool. However, Figure 3(b) reveals node crossing which a node occludes the other node and makes it impossible to see all the information at first glance. It's not in the paper, but wonder if 'over-mouse' feature is available to see the label.
The numbers in Figure 4(a) were thought to be ranks. It says they labeled each community with a unique integer but it would have been better if they used the rank of the hub as default. For example, in Figure 4(a), the unique number of the hub in each community are Pakistan-4, Bangladesh-2, India-8, Algeria-6 and so on. Not sure if the measure is different between Figure 1 and 4, but the rank order according to Figure1(a) is Algeria, Bangladesh, India, and Pakistan.
Overall, they have put a lot of effort in showing the graph in alternative ways. However, it is questionable whether it applies to general graphs. It seems the data needs to have a rank or order among the nodes. That said, it would be nice to see other examples applied with different data sets.