Automating Animated Transitions in Statistical Graphics

Abstract

People use statistical graphics (e.g., bar or line charts) to analyze data or convey insights from data. As data often involves many variables and aspects, it is inevitable to use more than one graphic. By transitioning among the charts, people relate findings across them and build up a holistic understanding of data. To facilitate this process, visualization researchers have studied effective methods for analyzing and conveying transitions, typically via animation. Indeed, animated transitions have been employed not only in research but also in public media, such as video news, slideshows, and interactive articles. Yet, authoring animations of transitions requires significant effort: authors need to 1) understand the semantics of changes across the charts and organize them in an easy-to- follow way, 2) consider many animation designs with varied timing and techniques (e.g., staging), and 3) implement animations by manually specifying low-level details, such as selecting visualization components and assigning animation parameters. I address these challenges by automating animation designs for a given transition. I have contributed empirical studies for a deeper understanding of animated transition designs (Aggregate Animations), formal representations and computational models for transitions (GraphScape) and animations (Gemini), and recommender systems for staged animation designs (Gemini, Gemini2). I evaluate the representations and models through user studiesand assess how they align with users’ preferences and perceptions.