The persuasive phase of visualization
Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining
| Hi-index | 0.00 |
Available automated information graphic generation systems create simple graphics by fitting data to predefined formats, based on data characteristics and task specification. However, beyond task and ordering of the data variables, current systems do not support the underlying messages that most information graphics have. These messages can often be conveyed through emphasis, which vary the visual properties of a graphic to direct the viewer's focus. We present a system, based on a fluid hierarchical model, which can automatically generate complex information graphics, not limited to predefined formats and capable of incorporating many emphasis techniques, as well as be modified without complete regeneration. Instead of classifying information graphics as disparate families of formats, our approach capitalizes on their fundamental similarities. We view different graphical formats as different perspectives of the data, a direct mapping from the data structure to the graphical structure. This graphical structure is constructed from a unifying element, the transformable coordinate space (TCS). A TCS is a locally Euclidean spatio-temporal object with its own coordinate system, a definite boundary (shape), and different visual properties (such as color and size) with which to encode data. TCSs can be nested to form a tree structure, by embedding one coordinate system into another. The commonly seen information graphics and their variations can all be described as nested TCS structures. Furthermore, as they are variations of the same structure, graphics that encode the same data can be incrementally transformed into each other. By unifying information graphics into structures built with one type of element, automated generation becomes a simple selection of the visual properties and their transformations. As each visual property is mapped separately to the data variables without being constrained to some predefined combinations, the results are not limited to pre-coded formats as in previous systems. In addition, different emphasis techniques, being variations of visual properties, can be easily added. Morpherspective, a prototype system using the TCS model, can automatically generate many different types of information graphics with minimal user input. Unique to Morpherspective is that the generated graphic can incorporate many different kinds of emphasis techniques, including user-defined ones.