Lee, B., Parr, C. S., Plaisant, C., Bederson, B. B., Veksler, V. D., Gray, W. D., & Kotfila, C. (2006). Treeplus: Interactive exploration of networks with enhanced tree layouts. Visualization and Computer Graphics, IEEE Transactions on, 12(6), 1414–1426.
Video demo of the interface available: http://www.cs.umd.edu/hcil/treeplus/TreePlusVideo/TreePlus.html
Code (.NET) is also available: http://www.cs.umd.edu/hcil/treeplus/sourcecode.htm
This interface design is intended to allow users to explore large graphs (ie. ontologies), expanding them as needed. Thus the user gets an overview, and can zoom in to sections of the graph for a closer look, following Shneiderman’s Visual Information-Seeking Mantra: “Overview first, zoom and filter, then details-on- demand.” Five design goals guided development:
- Take advantage of human perception of trees
– previous work showed tree visualizations are broadly applicable to many audiences
- Make as many nodes readable as possible
– supports users’ scanning behavior
- Maximize stability of layout
– changes to the interface are predictable (unlike force-directed graphs)
- Offer preview before committing
– supports information scent
- Provide multi-step animations so users can follow changes
– makes it possible for the user to follow what is changing in the interface
– the assumption is users will be most interested in the most recently opened node
The TreePlus interface was compared to a traditional graph interface in a controlled study. As the density of data increased, the advantages of the TreePlus interface increased.
2x2x6 (2 interfaces with 2 densities of the graphs by 6 tasks) repeated- measure, within-subject design.
Participants: 28 participants, CS and Engineering students able to quickly understand graph terminology. They already understood graph and spanning tree definitions.
1. Completion Time
2. Success Rate – percentage of tasks correctly answered
3. Error- difference between the correct answer and participant response
4. User Confidence – user questionairre
Participants first received training on the first interface and the eye tracking system was calibrated for them.
Each task included 2 practice trials and from 3 to 5 timed trials depending on the tasks. Participants were allowed to ask questions during the practice trials but not during the timed ones.
TreePlus performed significantly better for most of the tasks
– completed the tasks faster and with fewer errors
– reported higher levels of confidence in their answers
– most of them preferred TreePlus
“Results suggest that visualization and interaction techniques can effectively support incremental exploration of a graph, and can reveal the graph structure superimiposed onto a tree structure”
“I was very comfortable using it because I am used to the hierarchical structure” and “I think trees are logical and ordered arrangement of the graphs.”