• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

The ability to interact with (monitor and steer) a computation while it is running is a practical way of coming to understand it. The size of some computations can overwhelm both the interaction software and the end user. One technique for handling large computations is the use of exploratory visualization. It allows users to discover the computation's behavior by providing mechanisms for incrementally adjusting their perspective on the application, as their interests change over time. This paper presents a model for the interactions necessary to support exploration. Of particular interest is formal specification of operations for direct manipulation of the generated visualizations. The model describes how users can shift/enlarge/narrow their data collection, modify the set of visualizations used (possibly generating new visualizations onthe-fly), and provide feedback to the computation. The key idea behind the model is that of separating the information content of a visualization from the graphics used to represent the information. This makes it easier to define the semantics of direct manipulation, facilitates rapid initial construction and gradual refinement of custom visualizations. The paper presents the model, reports on the development of a prototype system, and provides illustrative examples. The prototype includes heuristic for generating visualizations at runtime and demonstrates the operations defined in the model.

Citations