Title: Automating Presentations of Data with Program Synthesis
Advisor: Rastislav Bodik and Alvin Cheung
Supervisory Committee: Rastislav Bodik (Co-Chair), Alvin Cheung (Co-Chair), Cecilia Aragon (GSR, HCDE), Jeffrey Heer, and Amy Ko (iSchool)
Abstract: While data presentation plays a crucial role in the data analysis pipeline, creating effective presentations (e.g., data visualizations or data navigation applications) from complex data is far from an easy task. In particular, in order to create an effective data presentation, users often needs to (1) query databases to obtain and combine data of interests, (2) understand functionality provided by existing data presentations libraries and prepare the input data to match the data shape expected by the libraries, (3) utilize knowledge to improve the effectiveness of the design. In fact, all these tasks assume programming knowledge from the user, and this knowledge barrier limits non-expert users’ ability to create expressive data presentations for their tasks.
Our work aims to automate the data presentation pipeline by automatically synthe- sizing programs that can query databases, prepare data, and present data from user demonstration. In this document, we will first review existing data presentation solutions and program synthesis techniques to motivate the design of next gener- ation synthesis-based data presentation solution. We will then review our prior work on (1) Scythe, a SQL query synthesizer that can synthesize SQL queries from small input-output examples provided by the user, and (2) Falx, a visualization by demonstration tool that automatically infers data preparation and visualization scripts from sketches of visualizations provided by the user, and (3) Draco, a visualization reasoning tool for optimizing and synthesizing visualizations from partial design specification. Finally, we will propose new approaches that will enable users to create more expressive data presentations with less demonstration effort. Concretely, we plan to design an opportunistic programming approach for automatically reusing and refining existing expert designs from users’ demonstra- tions of adaptation. We envision this approach would enable user to build more expressive data presentations without programming knowledge, and we plan to ground this approach in the domain of creating interactive visualization and mobile applications for data navigation.