| Recently, the amount of digital information available in the world has been growing at a tremendous rate. This huge, heterogeneous, and complicated data that we are continuously generating could be an incredible resource for us to seek insights and make informed decisions. For this knowledge extraction to be efficient, visual exploration of data is demanded in addition to fully automatic methods, because visual exploration can integrate the creativity, flexibility, and general experience of the human user into the sense-making process through interaction and visualization techniques.;Due to the scale and complexity of data, robust conclusions are usually formed by coordinating many sub-regions in an information space, which leads to the approach of multi-focus visual exploration that allows browsing different data segments with multiple views and perspectives simultaneously. While prior research has proposed a myriad of information visualization techniques, there still lacks comprehensive understanding about how visual exploration can be facilitated by multi-focus interactive visualizations. This dissertation investigates issues and techniques of multi-focus visual exploration through five design studies, touching various types of data in a range of application domains.;The first two design studies address the exploration of numerical data values. KronoMiner presents a multi-purpose visual tool for exploring time-series based on a dynamic radial hierarchy; and the ChronoLenses system supports exploratory visual analysis of time-series by allowing users to progressively construct advanced analytical pipelines. The third design study focuses on the exploration of logical data structures, and presents DAViewer that facilitates computational linguistics researchers to explore and compare rhetorical trees. The last two design studies consider the exploration of heterogeneous data attributes (or facets). TimeSlice facilitates the browsing of multi-faceted events timelines by organizing visual queries in a tree structure; and PivotSlice aids the mining of relationships in multi-attributed networks through a dynamic subdivision of data with customized semantics.;This dissertation ends with critical reflections and generalizations of the experiences obtained from the case studies. High-level design considerations, conceptual models, and visualization theories are distilled to inform researchers and practitioners in information visualization for devising effective multi-focus visual interfaces. |
