Jason Chuang received his Ph.D. in Computer Science from Stanford University, and is now a post-doctoral researcher at the University of Washington. He investigates how people work with data and each other to accomplish real-world analysis tasks. By examining how automated techniques and users' actions jointly contribute to the analytic process, he develops improved visualizations and analysis algorithms. His research draws on work from multiple disciplines including information visualization, human-computer interaction, machine learning, and natural language processing.
Title: Designing Visual Analysis Methods
Abstract: Scientific discoveries today are increasingly powered by analysis of massive datasets. As our unprecedented access to data continues to grow, how do we build analysis tools to support scientific breakthroughs of tomorrow?My research focuses on the design of interactive visualizations, statistical models, and integrated analysis workflows to enable people and algorithms to work in tandem to yield insights from complex data.In this talk, I first describe my experiences developing a variety of text analysis tools. I present guidelines for creating effective model-driven visualizations, and demonstrate that model design is just as critical as visual design in determining the effectiveness of a tool. I then examine the effective design of statistical models. I show that developing and deploying machine learning techniques can be a challenging analysis task in itself, which benefits from the application of visual analytics. Applying a human-centered iterative design method to statistical topic modeling, I contribute methods, tools, and frameworks that allow users to more efficiently utilize domain expertise to assess model outputs and explore modeling options. My approach improves our understanding of topic modeling techniques, and leads to tools and models that are responsive to user needs and support domain-specific applications.