GraphTrail

Abstract

Exploring large network datasets, such as scientific collaboration networks, is challenging because they often contain a large number of nodes and edges in several types and with multiple attributes. Analyses of such networks are often long and complex, and may require several sessions by multiple users. Therefore, it is often difficult for users to recall their own exploration history or share it with others. We introduce GraphTrail, an interactive visualization for analyzing networks through exploration of node and edge aggregates that captures users' interactions and integrates this history directly in the exploration workspace. To facilitate large network analysis, GraphTrail integrates aggregation with familiar charts, drag-and-drop interaction on a canvas, and a novel pivoting mechanism for transitioning between aggregates. Through a three-month field study with a team of archeologists and a qualitative lab study with ten users, we demonstrate the effectiveness of our design and the benefits of integrated exploration history, including analysis comprehension, insight discovery, and exploration recall.

Analyzing Organizational Routines in Online Knowledge Collaborations

“…Prior research investigating co-production dynamics as either aggregated sequence log to represent a particular process, creating multiple "snapshots" of the process over time, or "collapsed" sequence logs to calculate a pair-wise relationship between entities, forming a network of these entities and using social network analysis techniques to study the structure of the network [37,39,30]. Less relevant to our investigation are prior works that employed information visualization techniques --rather than a formal knowledge representation -to depict the dynamics of collaborative work [18,59,65].…”

“…In order to illustrate our proposed framework, we describe a case study that explores the co-production sequences. Building on theoretical frameworks of generative routines [18], our empirical analysis focuses on the sequences of contributors collaborating to create a Wikipedia article. Using a a representative sample of 93 articles from English Wikipedia, with 37,515 editing activities made by 16,616 distinct contributors, this case explores whether the co-production entails a small cohort of editors who work closely together, or if conversely the production of content within Wikipedia is built upon lengthy editing sessions by the same editor.…”

Expressive Authoring of Node-Link Diagrams With Graphies

“…GraphTrail [14] takes a radically different approach, focusing on the multi-variate data associated with the network's elements, that get displayed using charts that bin nodes and links according to selected attributes. Users can get an overview of the network by juxtaposing several such charts.…”

Juniper: A Tree+Table Approach to Multivariate Graph Visualization

“…(4) Layout adaption works by adjusting the layout so that a direct association between the nodes/edges and their attributes can be established. This is a broad category that includes placing the nodes in a scatterplot defined by two attributes as in GraphDice [3] or aggregating nodes into bar charts as in GraphTrail [6]. Another strategy is to linearize (parts of) a node-link layout so that it can be easily juxtaposed with a table visualizing node or edge attributes.…”