VisLink: Revealing Relationships Amongst Visualizations
Christopher Collins, Sheelagh Carpendale, Gerald Penn

Abstract
We have developed VisLink, a method by which visualizations and the relationships between them can be interactively explored. Our approach uses multiple 2D layouts, drawing each one in its own plane. These planes can then be placed and re-positioned in 3D space: side by side, in parallel, or in chosen placements that provide favoured views. Relationships, connections, and patterns between visualizations can be revealed and explored using a variety of interaction techniques including spreading activation and search filters.

We have also devised a formalism for understanding and comparing methods of mutli-relationship visualization, and analyze how the most popular methods (compound graphs, coordinated multiple views, Semantic Substrates) compare to VisLink. VisLink readily generalizes to support multiple visualizations, empowers inter-representational queries, and enables the reuse of the spatial variables, thus supporting efficient information encoding and providing for powerful visualization bridging.

Ongoing research is investigating the application of VisLink to real analysis scenarios in various data domains. We are also extending the capability for powerful inter-visualization queries.

As information visualizations continue to play a more frequent role in information analysis, the complexity of the queries for which we would like visual explanations also continues to grow. While creating visualizations of multi-variate data is a familiar challenge, the visual portrayal of two sets of relationships, one primary and one secondary, within a given visualization is relatively new. With VisLink, we extend this direction, making it possible to reveal relationships, patterns, and connections between two or more primary visualizations. VisLink enables reuse of the spatial visual variable, thus supporting efficient information encoding and providing for powerful visualization bridging which in turn allows inter-visualization queries. For example, consider a linguistic question such as whether the formal hierarchical structure as expressed through the IS-A relationships in WordNet is reflected by actual semantic similarity from usage statistics. This is best answered by propagating relationships between two visualizations: one a hierarchical view of WordNet IS-A relationships and the other a node clustering graph of semantic similarity relationships. Patterns within the inter-visualization relationships will reveal the similarities and differences in the two views of lexical organization.

VisLink supports the display of multiple 2D visualizations, each with its own use of spatial organization and each placed on its own interactive plane. These planes can be positioned and re-positioned supporting inter-visualization comparisons; however, it is VisLink’s capability for displaying inter-representational queries that is our main contribution. Propagating edges between visualizations can reveal patterns by taking advantage of the spatial structure of both visualizations. In this paper we will explain our new visualization technique in comparison to existing multi-relationship visualizations.

We have developed interaction methods for easy manipulation of VisLink planes in 3D, including a set of restricted-movement widgets:

We have also provided a set of default views of the 3D space, including (a) planes aligned side-by-side, (b) open "book view", (c) top view, (d) book view from the top, and (e) side view:

Through exploring the various view points, patterns in the inter-plane relationships can be discovered.

Interaction with the 2D planes in VisLink is always equivalent-to-2D. All mouse actions on a plane are transformed into plane-relative coordinates, and the underlying visualization is responsible for managing the mouse interaction. The consequence of this is that existing prefuse-based visualizations can be placed on a VisLink plane without modification to their rendering or interaction. Each plane can be swung forward for single-plane interaction in an equivalent-to-2D mode, then placed back into the 3D space:

We have applied VisLink to existing prefuse-based visualizations, incorporating a treemap of the occupations of congress people, their individual fundraising success, and the area of the U.S. they were elected to represent:

Future Directions

VisLink offers many areas for future research. We are investigating methods to reduce edge congestion, new application scenarios, methods to suggest appropriate plane ordering (inter-plane edges are only visible between adjacent planes), and more that we can't tell you about yet...

This work is an extension of the excellent prefuse information visualization toolkit.

[Download AVI]

Note: You may need the free TechSmith Screen Capture codec.