PESCE: A search-based approach to the automated generation of visual presentations

One of the most common means of making information more useful is by using graphic presentations. By taking advantage of the enormous visual bandwidth of humans, and their remarkable perceptual abilities, graphic presentations can organize information so that it is faster to process and easier to understand and remember. Unfortunately, the dynamic nature of information and the different characteristics of users trying to use it conspire to lessen the utility of static diagrams; dynamically generated graphic presentations are therefore highly desirable.; In recent years, there has been ample work focusing on the automated generation of graphic presentations. Several visualization systems have been developed, and in general, they share two limitations: the lack of user tailoring, and the absence of empirical validation for design rules used in generating presentations. The claim of this thesis is that an effective visualization can be automatically generated by using simple search techniques and generic visualization knowledge that can be empirically validated and tailored to users. We present a framework for the automated generation of presentations about complex information, based on a bottom-up, search-based constructive approach. This framework has been implemented in a prototype system called PESCE (""Presentation Engine for Software Comprehension and Explanation""). PESCE automatically selects an effective visualization by searching a space of possible designs to visualize complex information; it consists of a simple depth-first search with constraint checking and backtracking, and includes heuristics for search efficiency. The information to be visualized was at first limited to software artifacts, but later was extended to other domains that involve complex, relational information.; A series of experiments with human subjects were conducted to test the feasibility of the approach. The final results suggest that a tool like PESCE will be able to generate presentations that are reasonably close in quality to hand-made presentations, in a fraction of the time it takes a human author to design one.