Sketch-based Support for Expressive Personal Visualization

Thesis Supervisors

• Theophanis Tsandilas, Chargé de Recherche, Inria & Université Paris-Saclay
• Fanny Chevalier, Associate Professor of Computer Science and Statistics, University of Toronto

Jury Members

• Sheelagh Carpendale, Professor of Computing Science, Simon Fraser University , Rapporteur & Examinatrice
• Daniel Keefe, Professor of Computer Science, University of Minnesota, Rapporteur & Examinateur
• Petra Isenberg, Directrice de recherche, Inria & Université Paris-Saclay, Examinatrice
• Samuel Huron, Maître de conférences, Telecom Paris, Examinateur
• Emmanuel Pietriga, Directeur de recherche, Inria & Université Paris-Saclay, Examinateur

Abstract

Expressive personal visualization utilizes flexible visualization structures to make visualization personal, namely, to include personal data, incorporate the visualization context, enable a meaningful choice of aesthetics, retain personal visual style, or support personal subjective emphasis. Within this thesis I explore sketching-based interactions and workflows for supporting expressive personal visualization, specifically the attributes of personal visual style, subjective emphasis, and inclusion of context.

I begin by collecting a set of visualization sketches drawn by 8 people envisioning visualizations for personal data. I  triangularize this set with other examples of expressive personal visualization to outline potential research directions, identifying 2D and 3D temporal representations as well as custom glyphs as areas of interest.

Temporal representations, i.e. timelines, are essential for communicating chronological narratives, but existing visualization tools tend to support only conventional linear representations, failing to capture idiosyncratic conceptualizations of time. In response, I built Timesplines, an authoring tool that allows people to sketch multiple free-form temporal axes and populate them with heterogeneous time-oriented data via a lazy data binding workflow. Authors can bend, compress, and stretch temporal axes to subjectively emphasize time intervals. To contextualize the data they can annotate axes with text and figurative elements. The results of two user studies show how people appropriate the concepts in Timesplines to express their own conceptualization of time, and the curated gallery of images demonstrates the expressive potential of this approach.

To further explore subjective emphasis and contextualization I move from 2D to 3D, investigating personal immersive narratives. Immersive experiences can help bridge cultural divides through perspective taking, but typically must be authored by people with programming expertise. Moments is a system to author and share personal experiences, specifically focusing on experiences of migration, in an immersive environment. Authors can recreate personal contexts through surround imagery, 2D and 3D content, ambient audio, audio annotation, and narrative transitions enabled by linking scenes together with teleporters. The author integrates their subjective experience by emphasizing parts of the surround imagery via blurring, focusing, and warping. We demonstrate the support for emphasis and contextualization with a small gallery of authored narratives. Future work will evaluate how this application facilitates people in expressing experiences of migration. 

Finally, I focus on visualization workflows that retain personal visual style. Sketches automatically manifest personal visual style, but moving from a sketch to a full-fledged visualization often requires throwing away the original and recreating it from scratch. We aim to instead formalize sketches, enabling iteration and systematic data mapping through a sketch-based templating workflow. In this workflow, authors sketch a representative visualization and structure it into an expressive template for an envisioned or partial dataset. In order to demonstrate and evaluate the proposed workflow, I implement DataGarden, which is evaluated through a reproduction, a freeform study, and a visualization gallery. We discuss how DataGarden supports personal expression, identifying cases which demonstrate the limitations of our approach and discuss avenues for future work.

I conclude with a comparison of the evaluations and how the systems support both novice and expert users, outlining directions for future work on expressive personal visualization authoring. I further discuss the narrative aspect of the resulting galleries and how the visualizations might be considered through the lens of different narratives structures.