TwD98 Summary: Jonathan Meddes

I am a PhD student in the Department of Computer Science at the University of Edinburgh working with British Telecom (BT Laboratories in Matlesham) to research automated computer visualisation. At present the majority of our work is concentrated around constructing a visualisation framework; central to this is (1) the visualisation pipeline that processes raw data objects into a visual representation, (2) modelling the users interpretation of the visual representation and original data, and (3) integration of environmental factors such as user knowledge.

Our visualisation framework initially classifies the input data into a data network where appropriate data semantics, determined from user interviews, are captured in the relationships between data classes and objects. The data network is used in the first stage of the visualisation pipeline, subsequent stages of the pipeline use structure mapping rules to identify properties of the data network and map them onto suitable visual counterparts. To aid comprehension we want to retain the data semantics in the visual representation and, where appropriate, the mapping rules must respect the visual information encoded in the data (e.g., perceptual groupings). Although an aesthetically pleasing representation is desirable, structure mapping must ensure that the visual representation is a syntactically and semantically valid representation of the original data. The definition and integration of the structure mapping rules into our visualisation framework is the focus of our research.

Type of Diagrams Studied and Representation Medium

We are interested in effective visual representations of highly integrated topological, geographical and quantitative data objects. The visual representations supported by our framework are exact geographical representations, approximate geographical representations and topological representations (i.e., network diagrams). The final stage of the visualisation pipeline renders the visual representation on a graphical computer display; the framework supports active objects that permit human-computer interaction with graphic (and therefore, data) elements of the display.

 

Accompanying Non Diagrammatic Representations

The underlying relational database stores data in textual tables; each column contains data items from the same domain and each row contains related data items. An effective visualisation improves our ability to interpret large volumes of data contained in the tables; naturally, the two representations are closely related and can be used together, but the nature of visualisation applications means this is rarely the case.

 

Application Domain and Approach

We are working with a group of users (currently, our 'users' are both the owners and consumers of the data) that manually create visual representations from original data. The goal of our research is not to faithfully replicate the behaviour of an 'expert' user, but to identify methods that create effective visual representations that respect user domain knowledge while not being constrained by user prejudice. We hope user trials will show this facilitates alternative 'creative' interpretations of the data.

For further information contact Jonathan Meddes <jmx@dcs.ed.ac.uk>