Proactive Analytic Workspaces fro Heterogeneous Data Frank Shipman Dept. of Computer Science and Engineering Center for the Study of Digital Libraries
Vision: Application of User/Interest Modeling to Support Data Analysis Domain Analysts Heterogeneous Data Repository Proactive Analytic Workspace Access and Selection Interpretive Activity Data Similarity Assessments Interest Models
Per. Con: A Personal Data Analysis Environment Prototype Repository Viewer Visual Workspace Suggestion History
Evaluation of Initial Capabilities • 24 participants performed tasks based on two years of weather and river data • Results – Availability of a persistent workspace increases the number of data objects investigated by users and improves the users’ perception of the environment – Recommendation effects are muted without the workspace – Activity moved from the repository viewer to the workspace when available – Many users valued the recommendations and actively sought more recommendations
Lots of Opportunities for Collaboration • Our focus is on proactively supporting human analysis • We are domain agnostic and can modify the tools to work with most types of data • Our architecture includes ability to add/replace data types, feature extractors, similarity measures, etc.