Content area
As Tufte emphasizes, “above all else, show the data”. Consistently highlighting key aspects of the data in visualizations ensures a more reliable, quick, and insightful understanding of data, particularly when interpreting various aspects of the data. This is particularly applicable for hierarchical data, which have many facets of relational information. A hierarchy is a system that organizes entities by rank, order, or importance, using parent-child relationships at its foundation. These relationships often result in derivation of various other kinds of relationships within the hierarchy, some of which are widely applicable and considered of primary importance across domains, while others are domain-specific and of special importance. Examples of these relationships include siblings, levels and paths. The nodes and edges that comprise these relationships are often collectively referred to as topologies by the visualization community. In our research, we further distinguish the topologies of single node and edge elements, referred to as basic topologies and those formed by sets of nodes and edges, which we term composite topologies. Having the ability to explicitly represent composite topologies and provide support for interactive operations on them is useful for revealing patterns, facilitating analysis, and drawing meaningful conclusions from hierarchical data.
Tree visualizations are a subclass of hierarchical visualizations designed to represent data with strict parent-child relationships in a tree structure. These techniques emphasize visual clarity in conveying hierarchical depth and structural relationships. Existing tree visualization techniques and systems primarily focus on representing node entities and parent-child relationships between node entities, and to some extent supporting operations on nodes and edges. These visualizations often reveal various other composite topologies incidentally through their design. Some relationships, like parent-child relationships, are easier to interpret in these visualizations than composite topologies because they are explicitly represented. Other relationships represented by composite topologies require deliberate effort by users to interpret their structure from what they see. Explicit representation of composite topologies can allow users to see those topologies directly rather than discerning them from their constituent node and edge representations, opening new avenues for interactive data exploration and analysis of hierarchies.
This research explores how existing tree visualizations represent information and how these visualizations can be extended to support representation of composite topologies and interactive operations on them. We examine the concept of topologies: the structures in a hierarchy, comprising nodes and edges, that constitute various relations. We develop a software architecture and a data pipeline to support tree visualization design, along with an implemented tree visualization system called Hieros. We conduct a user evaluation to study the usability and utility of representing various relations and interacting with them in tree visualizations.
This dissertation makes five main research contributions. The first contribution is C4D3, a JavaScript-based coordination library for building coordinated multiple view visualizations using the D3 library for web-based visualizations. It is based on the Live Properties architecture of the Improvise visualization system. C4D3 aims to adapt and implement the coordination architecture of Improvise as an independent library that can be used with web-based visualization libraries such as D3, and eventually make it possible to build and coordinate tree visualizations via their topologies.
The second contribution is a conceptual model of representative suitability, which is used to assess the capability and suitability of tree visualizations to represent the structures and their attributes in hierarchical data. The model aims to help in predicting how likely tree visualizations are to effectively visually represent specific topologies and the information associated with them.
The third contribution is the design of Hieros, a tree visualization library with a modular software architecture and a hierarchical data pipeline. The architecture supports explicit representation of topologies and interactive operations on them at various stages of the visualization pipeline.
The fourth contribution is a reference implementation of Hieros, a tree visualization library implemented in the Improvise visualization environment to support construction of topology-augmented tree visualizations.
The fifth contribution is a user evaluation to assess the utility and usability of tree visualizations in explicitly representing topologies.
Overall, this dissertation contributes to the systematic understanding of tree visualizations in terms of topological structures that form the building blocks of hierarchies, the building of a visualization system and data processing pipeline to support topology-centric tree visualizations, and demonstration of their application to a variety of knowledge domains.
