Marian Dörk
PhD Thesis

Visualization for Search:
Exploring Complex and Dynamic Information Spaces

Department of Computer Science, University of Calgary, July 2012
Advisors: Sheelagh Carpendale and Carey Williamson

Abstract

Digital information has become immensely abundant and ubiquitous in our lives. During work or leisure, we traverse search results, news items, and status updates for a wide range of purposes. Typically, information is displayed as lists ordered by relevance or recency, which may be sufficient for displaying a few items. However, considering the substantial increases in complexity and scale of many information spaces, lists are of limited use for discerning how resources are related or what a collection contains. In contrast, information visualization can utilize our perception to reveal patterns and provide overviews of large datasets.

This dissertation investigates the potential of visualization for navigating growing information spaces, and presents methods for arranging and accessing digital information. Unlike most information seeking models that emphasize informational deficiencies, the approach taken here highlights information practices characterized by curiosity. Picturing the searcher as an information flaneur with an exploratory attitude, we can envision innovative interfaces for accessing information.

Three design studies are presented that are aimed to support casual exploration while attending to the challenges and opportunities of informational abundance. Addressing the diversity of relationships in many datasets, EdgeMaps combines spatialization with graph-drawing techniques. Using an integrated representation it is possible to see the semantic or temporal extent of influence, for example, among philosophers or musicians. Visual Backchannel is motivated by highly dynamic communication channels during large events and integrates evolving visualizations for topics, participants, and photos of ongoing conversations. Approaching the difference in scale between large result sets and individual items, Fluid Views narrows the gap between overview and detail in a zoomable search environment, in which results portray similarity via position and relevance via size and level of detail.

The research is validated by prototype implementations, a large-scale web study, deployment during conferences, and consultations with search experts. The dissertation ends with a critical reflection about the role of values in information visualization. The insights gained from empirical and theoretical methods suggest novel forms of information seeking and raise challenges for research and design.

Publications on PhD Topic

Journal Articles

Conference Papers

Workshop Contributions

Other Contributions

Downloads

Design Studies