Optimizing readability for complex UML diagrams generation

Abstract

Information visualization provides a way to convey key notions and concepts in a single image. It is a powerful tool if the visualization product is readable and has the appropriate level of information for the targeted audience. Readability means the ease with which the graphical representation is understood and processed by the viewer. Although several visualization techniques exist, this thesis work focuses on UML diagrams for software architectures’ representation. Tools exist for automatic UML diagram generation from software architecture description but the outputs for complex diagrams are cluttered and unreadable. The aim of this thesis is to provide a solution to generate automatically readable software architecture visualizations for a computer scientist’s audience of all level. It is an effort to make progress in this research field by applying theoretical state-of-the-art research findings to a design study and evaluate how they perform. This work has been conducted for Ericsson (telecommunications Company) with the goal of improving the communication within and between teams as well as keeping their documentation updated in a simple manner. The result of this thesis work is a prototype enabling the users to control several abstraction parameters and control diagram readability depending on the targeted audience. Experimentation and surveys have been conducted to determine the tool’s best settings for the audience targeted during this study. The prototype has been developed in Java and uses PlantUML as its UML diagram visualization engine.