A dual-view approach to interactive network visualization

Abstract

Visualizing network data, from tree structures to arbitrarily connected graphs, is a difficult problem in information visualization. A large part of the problem is that in network data, users not only have to visualize the attributes specific to each data item, but also the links specifying how those items are connected to each other. Past approaches to resolving these difficulties focus on zooming, clustering, filtering and applying various methods of laying out nodes and edges. Such approaches, however, focus only on optimizing a network visualization in a single view, limiting the amount of information that can be shown and explored in parallel. Moreover, past approaches do not allow users to cross reference different subsets or aspects of large, complex networks. In this paper, we propose an approach to these limitations using multiple coordinated views of a given network. To illustrate our approach, we implement a tool called DualNet and evaluate the tool with a case study using an email communication network. We show how using multiple coordinated views improves navigation and provides insight into large networks with multiple node and link properties and types.

Multiple Views: different meanings and collocated words

“…The general idea to put two views close together is a simple yet effective design strategy, and consequently there are many words that have the same sense, including sideby-side, dual views [PW03], and parallel views (see Table 7 for a summary of terms, frequencies and examples from the v6Y corpus). There are many examples that use this technique, including Henry and Fekete [HdF06] and Namata et al [NSGS07], who demonstrate tools that display side-by-side network graphs. The term parallel is sometimes used to explain several side-byside views, or bracketed views "where a principal view is supported with two [or more] additional views from slightly different parametrisations" [Rob04].…”

MostVis: An Interactive Visualization Supporting Automotive Engineers in MOST Catalog Exploration

“…It has been found in many scenarios, that MCV improve user performance, e.g., [8,16,17]. However, it is important to design MCV applications very carefully to reduce unnecessary context switching and efficiently use display space [1,6,7].…”

RelEx: Visualization for Actively Changing Overlay Network Specifications

“…Abstractly, these domain problems map to the generic tasks of identifying and describing clusters and specific types of nodes including highdegree nodes, bridge nodes, and outliers. These generic tasks also occur in the visual analysis of email communication networks [24] and academic co-authorship [30].…”

“…The Honeycomb system from van Ham et al [45] and the work of Perer et al [35,34] also fit this mold. Design studies focused on email communication networks [24] and academic paper co-authorship [30] also use similar abstractions.…”