Graduate student, Computer Science Dept., UC Santa Barbara

John Roberts is a doctoral student in Computer Science at UC Santa Barbara who currently holds a fellowship in the National Science Foundation IGERT (Integrative Graduate Education and Research Traineeship program) at UCSB on Interactive Digital Media. His current research focuses on providing support for penbased interaction in video segmentation. Additional research interests include Human-Computer Interaction (HCI), especially as it intersects with Personal Information Management, and Ubiquitous Computing. Roberts took his M.S. in Computer Science from San Francisco State U. in 2005 with a thesis titled “Browsing in Large, Time-Dependent Data Set” (pdf). His co-authored papers (published or forthcoming) include: “Visualizations for Browsing in Large Datasets‿ (pdf); “Making Favorites Useful‿ (pdf); “Easy and Effective Virtual Tours on the World Wide Web‿ (The Society for Imaging Science and Technology, Internet Imaging VI. San Jose, CA, USA, January 18-20, 2005); “An Interface Markup Language for Web3D‿ (pdf); and “Histogram-Based Visualizations for Large, Time-Dependent Data Sets‿ (pdf).

Links: Home page | UCSB Four Eyes Lab

Research Sample: Excerpt from “Visualizations for Browsing in Large Data Sets”

The specific goal of our research is to determine the effectiveness of visualizations we create in allowing people to perceive patterns in personal data sets and to retrieve specific information from personal data sets. We have conducted a user trial demonstrating that, while both histograms and bargrams are both usable and useful, histogram visualizations are more effective than bargram visualizations for file retrieval in personal data sets. All visualizations are similarly effective in exposing patterns in personal data sets to users.

Our motivation for the use of histogram variants is simple. Users engage in various activities over the course of their life. Each activity is likely to generate a pattern of data in an automatically captured personal data store. As in any visualization task, we seek to expose patterns in the data to a user. If we can expose the patterns that exist in a personal data set to a user, the user can correlate that pattern with his or her life, and use an understanding of the patterns present to guide search and retrieval through browsing. Histogram variants are an effective representation of frequency over time, thus allowing individuals to perceive patterns in their personal data store. We make these visualizations zoomable and browsable, thus allowing users to interact with the data set to locate a specific piece of information they seek.