We present a new aggregation and multi-mode linked selection framework for data exploration. To enable scalable data overviews, aggregates group records by their attribute values and measure group characteristics within data summaries. To reveal details, linked selections visualize data distributions on aggregations upon interaction with three complementary modes: highlighting, filtering, comparison. This model is domain independent, expressive, minimal, and scalable, and constructs an exploration space without the complexity of manual visualizations and interaction specification tasks. We implemented this framework for tabular data as a web-based tool, Keshif. A Keshif data browser combines summarized aggregations on existing or calculated attributes, and individual records. Data exploration is supported from importing raw data, to authoring, sharing, and forking data browsers, through a fluid, consistent, rapid, and animated interaction design. We demonstrate aggregation designs for multiple data types (categorical, set-typed, numeric, timestamp, spatial) using various glyphs and non-overlapping visualizations (bar, line, icon, disc, geo-area). We illustrate examples from 130+ publicly published Keshif data browsers from diverse domains.
M. Adil Yalcin, is a Ph.D. candidate at the Department of Computer Science at University of Maryland, College Park, and a member of Human Computer Interaction Laboratory (HCIL). His goal is to lower human-centered barriers to data exploration and presentation. His research focuses on information visualization and interaction design, implementation, and evaluation. He is the developer of keshif, a web-based tool for rapid exploration of structured datasets. In his previous work, he developed computer graphics techniques and applications.