Tuesday, November 15, 2011

Reaction: Task taxonomy for graph visualization

The paper introduces an overview of low-level and high-level tasks that a lot of people would ask themselves. Even though the paper categorizes their individual questions and tasks which gives an impression of lacking generality, it still initiates to think about 'how-to-categorize' the tasks that people always perform with graphs.
Regarding to the definitions, I wonder if people use 'group' as a general meaning of a set of related nodes which is composed of several clusters as described in the paper. According to the description of correlation by Amar et al.'s low-level visual analytic task, it is to determine useful relationships betweeen the values. Regarding to the correlate task, the paper describes that "in the field of Information Visualization, the intended meaning of 'correlate' is to identify possible correlations". I guess they say this because in information visualization, being "useful" depends on the domain(or object/task) and the user's perspective, or maybe other criteria. Now, I wonder if correlate is a low-level task at all. The verbs used to describe the tasks by Amar et al. are find, compute, rank, characterize, identify, and determine. I personally think 'identify' and 'determine' are operations that involves decision making process which needs to set up criteria which may not be a low-level task.
It's good to know some differences between tasks. 'Find a path' and 'follow a path' are different since finding involves 'filter on links' while following involves 'scan on nodes'.
The authors could have done better elaborating the high level tasks. The questions look so random and seem they don't cover all the possible tasks. In their conclusion, they say they demonstrated all the complex tasks to series of low-level tasks but I wonder if they have described "all" the complex tasks.
By looking at Table 2, does it mean there cannot be just one visualization tool that can perform all the tasks? So, the visualization tool always has its domain, object, and task of different level which it focuses?