Spatializing visual exploration: Transforming interactive visual analysis into spatial representations to aid sensemaking

Waqas Javed, Purdue University

Abstract

Interactive visual analysis is the process of collecting insights about a dataset, while using one or more interactive data visualization techniques. The process is based on the progressive refinement of an analyst's knowledge about the dataset, while using different interaction and analytical operations. In the presence of large and complex datasets, interactive visual analysis often places a high cognitive load on the analyst in terms of memory, perception, and reasoning. For example, when studying a complex dataset such as U.S. Census demographics, scientific co-authorship networks, or sales data, an analyst may have to simultaneously keep track of current insights, potential avenues for further investigation, and also the steps taken to reach a particular result. In this work, we propose a new methodology for automatically spatializing the individual steps, involved in an interactive visual analysis, onto a visual canvas allowing users to easily recall, reflect, and assess their exploration. In essence, instead of using a single viewport with a set of visual representations that change as the user selects, filters, and navigates in the data, we cause different interactions with visualization to spawn a new visualization where the desired change—such as a filter, annotation, or change of visualization technique—has been enacted. The benefit of this radical approach to spatializing individual steps is to externalize not just the data being visualized, but the exploration process itself used to analyze the data, and hence to aid sensemaking. To validate this new methodology, and to measure its efficiency for different types of datasets and visualization techniques, we develop various novel information visualization systems. In particular, we present the StackZoom application for performing multifocus exploration in skewed-aspect visual spaces, PolyZoom for achieving multiscale multifocus exploration in 2D visual spaces such as Google Maps, and the ExPlates technique for analyzing large multidimensional datasets in web browsers. We also discuss different experimental studies, performed to investigate the performance of these applications for various practical tasks such as navigation, search, and comparison. Results from the experiments confirm the effectiveness of the new applications and in general highlight the usefulness of spatializing interactive visual exploration.

Degree

Ph.D.

Advisors

Elmqvist, Purdue University.

Subject Area

Computer Engineering|Electrical engineering|Computer science

COinS