you're reading...
HCI, InfoVis

Response to [Network Visualization …] by Shneiderman et al.

This paper offers a strategy for improving existing network visualization: 1) the layouts of the network is based on user-defined substrates, which means non-overlapping regions where node placement is based on node attributes and 2) users can interact to change the visibility of links thus preventing clutters and ensuring comprehensibility.

* Identifying tasks when viewing network visualization

To frame the goal of their visualization, the authors firstly identify 14 tasks that are relevant to at least 6 *challenges*. The 6 challenges are:

  • C1 Basic networks
  • C2 Node labels
  • C3 Link labels
  • C4 Directed networks
  • C5 Node attributes
  • C6 Link attributes

Based on these challenges, the authors propose

  1. 8 tasks for C1
  2. 2 more tasks for C[2, 3]
  3. 4 more tasks for C[5, 6]

* Interaction requirements

Dynamic queries technique is considered in their work (in my opinion of less originality)

* Semantic substrates for node layout

Given a representation of network, we can firstly semantically define (2 – 3) non-overlapping regions. For example, given a citation relationship network between papers, we can firstly place these papers into four categories: journal, conference, book and web. Following that, if there are N substrates, there will be N kinds of nodes and C(2,N) + N kinds of links for non-directed ones and A(2,N) + 2N for directed ones. And all these varieties are semantically-salient.

The authors point out two advantages of doing such substrates:

  1. Proportionally-sized regions would immediately give users some idea of the relative cardinality of each category;
  2. Users can quickly distinguish links that cross from one category (region) to another.

Also there might be some critiques toward using this method. Firstly it assumes the network inherently contains categorical attributes which might not be the case. Secondly by separating nodes and links, the linking might be made more complicated. And one other critique the authors fail to mention is scalability. What happens if there are 5, 10, 20 or 100 categories?

* The legal precedent example

In this example, there are three clear categories: supreme court (S) , circuit court (C) and district court (D). And citations between courts are distinctly categorized into external ones namely S2S, C2S, D2S, D2C as well as internal ones namely S2S, C2C and D2D. Users can dynamically query with certain criteria so that the visualization only shows part of the network hence much less likelihood of visual clutter.

However, there are still links cluttering which should be addressed in ways other than dynamic queries.


About Xiang 'Anthony' Chen

Making an Impact in Your Life


No comments yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Twitter Updates

%d bloggers like this: