| Information Visualization (Infovis) graphically encodes information to help a user explore a data set visually and interactively. This graphical encoding can take the form of widespread visualizations such as bar charts and scatterplots. Multiple visualizations can share the same functional space to form complete tools for visual exploration or for communicating information. There is multiple ways of combining these visualizations. The assembly of multiple visualizations can give some complex assemblies sometimes called hybrid visualizations.;A hybrid visualization is the result of assembling multiple simpler visualizations. For example, NodeTrix (Henry et al., 2007a) is composed of a node-link diagram and an adjacency matrix, and MatLink (Henry and Fekete, 2007a) adds arc links to an adjacency matrix. This integration of multiple visualizations can be a way to combine their advantages into a coherent structure. The integration can be achieved, for example, through color coding, or through explicit linking (such as with arrows), or through interaction (such as when different visualizations respond to the manipulation of others). Recent literature contains several examples of new hybrid visualizations, most often to deal with complex datasets where the user can benefit from multiple, complementary visual encodings of the same data. However, to date, there is almost no theory or framework to help researchers understand and characterize existing hybrids or design new ones.;This thesis advances the state of the art in hybrid visualizations in two ways: first, by developing a framework that defines and characterizes hybrid visualizations to help better identify, describe and design them, and second, by demonstrating a variety of novel hybrids. The hybrid visualizations we explored cover a wide range of possibilities. Two of the most general and widely used data types in Infovis, multidimensional multivariate data and graph (i.e., network) data, are each the subject of a chapter in the thesis, with novel hybrid visualization techniques presented for each. A wide range of possibilities for integration is also presented using a pipeline model.;After some preliminary material, chapter 2 of the thesis presents a conceptual framework that defines and characterizes hybrid visualizations. This framework was itself derived from experience designing the hybrid visualizations presented in the subsequent chapters. A hybrid visualization is described as a graphical encoding using other visualizations as building blocks. We present a pipeline to illustrate the assembly of a visualization, starting from the generation of basic shapes or glyphs, then placed on a layout, embellished by adding other graphical elements, then sent to some view transform operators and assembled on the same space. Simple charts can be described with this pipeline as well as more complex assembly and new hybrids are described. Chapter 3 presents ConnectedCharts, an example of a hybrid assembled on the assembly level of the pipeline, made of multiple multidimensional and multivariate charts explicitly connected by lines or curves showing the relationship between their elements. A user interface enables the interactive assembly of ConnectedCharts, including a wide range of previously-published hybrid visualizations, as well as novel hybrid arrangements. ConnectedCharts serve as an illustration of the conceptual framework in chapter 2, by exploring possible connections between different graphics depending on the relationship of their encoded data types. Chapter 4 presents another user interface, this time for graph exploration, that incorporates several highly integrated hybrid visualizations. A Parallel Scatter Plot Matrix (P-SPLOM) is presented that constitutes a fusion of a Scatter Plot Matrix (SPLOM) and a Parallel Coordinates Plot (PCP). A radial menu called the FlowVizMenu enables the modification of a visualization integrated at the center of the menu. This menu is also used to select the dimensions for configuring a third hybrid based on an Attribute-Driven Layout (ADL) that combines a nodelink diagram and a scatterplot. The characterization of hybrid visualizations offered by the conceptual framework, as well as the illustration of the framework by innovative hybrid visualizations, are the main contributions of this thesis to the Infovis community. |
