A Multi-layer Event Visualization for Exploring User Search Patterns in Literature Discovery with PUREsuggest

Understanding user behavior is at the heart of user interface design, but can only be quantified to some extent. Qualitatively analyzing individual usage sessions is especially important in open-ended tasks like literature search. In this paper, we present a visual representation of logging data that provides the basis for an in-depth analysis and annotation of search and exploration sessions. The visualization was developed to evaluate the citation-based literature discovery tool PUREsuggest and is thus aimed at visualizing logging data of a literature search system. Events are represented on a timeline in different layers as bars and icon-based glyphs, and contextualized by the additional visualization of item states and active user-set search modifiers such as filters or keywords throughout the session. We demonstrate the applicability of the visualization by evaluating excerpts of two user sessions as an example.