Expression of Interest

Overview

Expression of Interest is a project of the Visualization group of the faculty of Mathematics and Computer Science, Technische Universiteit Eindhoven. The project is supported by the VIEW programme of the Netherlands Organisation for Scientific Research (NWO) under research grant no. 643.100.502. The complete project proposal can be found elsewhere here [1], on this page an overview is given. The project has started May 2006, and is carried out by two PhD students.

The Visualization group has developed a number of techniques for data visualization. On our webpages you can find examples [2], also, have a look at SequoiaView [3] to see how visual realism can help in understanding data. The aim of the EOI project is to make a next step, and to develop more general insights and approaches. One small example where we think to have made such a step, is our work on zooming and panning [4, 5].

Data visualization concerns the visual presentation of data such that users can easily obtain insight [6]. A key ingredient for successful visualization systems is that they show interesting aspects in  a clear way. This is obvious, but there is little insight how this can be achieved. The aim of the project is to develop generic models and guidelines for this, which form a basis for a number of new methods and techniques.

Interest

The first key question is how users can be enabled to specify their (current) interest in an effective and efficient way. We propose to attach to each data element a Degree of Interest (DOI), to be specified by the user and to be used in later stages of the visualization pipeline.
Many different ways can be used to specify the DOI. A simple one is binary selection: The user picks or outlines interesting objects, shown in the visualization. Another approach is to enable the user to specify predicates on data values, for instance, by selecting a range of values for a parameter. More complex methods, such as Boolean expressions, can be used, but in practice these are often too complex. The main challenge here is to come up with powerful methods for the support of the user in the exploration of the data, that are easy to understand and to control.

Presentation

The second key question is how this user-defined interest can be used such that the visual attention for each element displayed corresponds to its importance. Graphic representations of data objects have attributes like color, shape, size, position, orientation, and texture. A simple approach to let interesting objects stand out is to parametrize a graphical attribute that is not used for other purposes to the DOI. For instance, if all objects are black, color interesting objects red; if all lines are thin, make interesting lines thick.
One can also add graphical elements, for instance by encircling interesting objects, or by assigning the background locally a different color. For simple cases, when all objects are homogeneous or when a single, static visualization is designed, these approaches work well. But for more complex situations  more powerful approaches are needed, which take full advantage of the capabilities of the human visual system.

Software Visualization

In the third place, visualization is an interactive and explorative process. How can we enable the user to change his interest continuously, and thereby to navigate through large data spaces efficiently and effectively?
New methods will be applied and tested in the context of software visualization. Software visualization is a challenging field, where large and complex data-sets have to be handled, and where multiple views and multiple visualization methods have to be used. During the analysis of software systems, the interest of the user will change quickly, and also, just a small part of the data will be of interest. This urges a need for visualization methods where interest is given an explicit and central role.

Support for Analytical Reasoning

Yedendra Shrinivasan focuses on support of the user while performing analytical reasoning. An important ingredient is to enable users to capture interesting aspects. This has led to a prototype system called Aruvi [7]. More information can be found here .

Evaluation and guidelines

Jing Li focuses on the development of better guidelines for information visualization, such that it can be predicted how much attention each items gets. Evaluation and user testing is important in this respect. As a first study, the effectiveness of scatterplots and parallel coordinate plots for judging correlation has been studied [8].

References

  1. J.J. van Wijk. Expression of Interest. Project proposal NWO, 2005.
  2. www.win.tue.nl/vis
  3. www.win.tue.nl/sequoiaview
  4. J.J. van Wijk and W.A.A. Nuy. Smooth and Efficient Zooming and Panning. In T. Munzner, S. North (eds.), Proceedings IEEE Symposium on Information Visualization (InfoVis'2003), IEEE Computer Society Press, October 2003, p. 15-22. Best paper award.
  5. Wijk, J.J. van, Wim A.A. Nuij. A Model for Smooth Viewing and Navigation of Large 2D Information Spaces. IEEE Transactions on Visualization and Computer Graphics, vol 10 no 4, July-August 2004, p. 447-458.
  6. J.J. van Wijk. The Value of Visualization. In C. Silva, E. Gröller, and H. Rushmeier (eds.), Proceedings IEEE Visualization 2005.
  7. Y.B. Shrinivasan Y.B, and J.J. van Wijk. Supporting the Analytical Reasoning Process in Information Visualization, to appear in ACM Human Factors in Computing Systems (CHI), 2008.
  8. J. Li, J.-B. Martens, J.J. van Wijk. Judging Correlation from Scatterplots and Parallel Coordinate Plots. Submitted to Information Visualization