About me

I am now working on the new IBM Watson team on information visualization, visual analytics, and cognitive computing in Cambridge, MA. You can check out my new homepage, though my publication list is still here. My new email is 'cdunne at us.ibm.com'.

In 2013 I graduated with my PhD in Computer Science from the University of Maryland, College Park. My interests lie primarily in human computer interaction, specifically information visualization and graph/network drawing (my portfolio). My advisor was Ben Shneiderman and I was a member of the UMD Human-Computer Interaction Lab (HCIL).

Projects

NodeXL: Network Overview, Discovery and Exploration for Excel

Network data structures have been used extensively in recent years for modeling entities and their ties for many diverse disciplines. Analyzing networks involves understanding the complex relationships between entities as well as any attributes, statistics, or groupings associated with them. Many specialized analysis tools and languages exist but are designed for expert practitioners with substantial technical skills. Our free and open source tool NodeXL makes network analysis easily accessible to non-programmers by integrating with the Excel spreadsheet application. With NodeXL, network visualizations become as easy to create as pie charts, statistic or metric rankings are available at the press of a button, and users can easily collect and map social media conversations around topics that matter to them.

GraphTrail: Analyzing large multivariate, heterogeneous networks with integrated exploration history

Large networks can be challenging to analyze because they can contain many nodes and edges in several types and with multiple attributes. Analyses of such networks are long and complex with several sessions and multiple users, making recalling or reproducing the exploration difficult. To address this, we designed GraphTrail: an interactive visualization for analyzing networks via node and edge aggregation. Aggregates are explored using familiar charts, drag-and-drop interactions, and a novel pivot mechanism. Users' interactions are automatically captured and this history is integrated directly in the exploration workspace. This design has been proven effective for analyzing large networks with many node and edge types, and the integrated history improves exploration recall and sharing of analyses with others.

Action Science Explorer: Enabling rapid understanding of scientific literature collections

Keeping up with rapidly growing research fields requires substantial effort for scientists, program managers, or venture capital investors. Current theories and tools are directed at finding a paper or website, not gaining an understanding of the key papers, authors, controversies, and hypotheses. We created Action Science Explorer (ASE) to help users rapidly generate easily consumable summaries of scientific literature. ASE uses link mining to create a citation network for a field and context for each citation, automatic clustering and text summarization techniques to extract key points, and potent network analysis, statistics, and visualization tools to aid in the exploration. ASE ties these tools together in a multiple coordinated window environment. ASE is featured as an NSF Discoveries Report.

STICK: Analyzing trends in science & technology innovation

The Science of Science and Innovation Policy strives to understand the relationships behind innovation and how ties between organizations, individuals, and funding sources affect growth. Our STICK project's goal is to overcome any bias towards popular or successful innovations by providing the much needed data and tools for analyzing innovations of all possible outcomes. We integrate data collection tools for trade press articles, scientific literature, patents, and blogs; text processing and crowd-sourcing to automatically identify people, organizations, and products; and network analysis and visualization to highlight key trends, relationships, groups, and outliers. This set of tools can be used to help students, scientists, and science and technology policy makers to understand and advance the process of discovery.

NetVisia: Heat map and matrix visualization of dynamic social networks

Visualizations of static networks in the form of node-link diagrams have evolved rapidly, though researchers are still grappling with how best to show evolution of nodes over time in these diagrams. One approach is NetVisia, a social network visualization system we designed to support users in exploring temporal evolution in networks by using heat maps to display node attribute changes over time. NetVisia's novel contributions to network visualizations are to (1) cluster nodes in the heat map by similar metric values instead of by topological similarity, and (2) align nodes in the heat map by events such as their first appearance. This clustered heat map approach is effective at detecting outlier nodes and time periods, as well as comparing the evolution of node attributes or statistics.

Online readability metrics for SocialAction

The omnipresent node-link visualization excels at showing network topology and features simultaneously, but frequently are not easily readable or difficult to extract meaning from because of inherent network complexity or size. Defining practical readability measures accelerates progress towards improved node-link visualizations, because it guides analysts in making both manual and automatic changes that improve quality. We have developed several novel global and local readability metrics that can be computed in real-time, and incorporated them into the network analysis tool SocialAction to provide real-time feedback to users as they manipulate the layout. With this metric feedback, users are able to localize areas needing improvement and manipulate the layout to create more effective visualizations.

MedCommix: A tool for aggregating and visualizing electronic health records

Electronic health records are being collected by health organizations every day. Designing a visualization to explore these huge databases is daunting due the limits imposed by screen space and data processing time. We developed MedCommix to visualize health data by focusing on similarities within clusters of patient records. We use a novel similarity measure to cluster patients by sentinel events in their history and bins of interesting events on either side. To deal with big data, not all records are clustered simultaneously. Instead, records are streamed from the database and run through an online k-median clustering algorithm that updates the results as information arrives. With a random stream, the results converge rapidly and users do not have to wait for an entire query to return before deriving insights.

NetGrok: Visualizing real-time network resource usage

Network administrators typically scan textual router logs looking for attackers or configuration issues. Unfortunately, these tasks have long been unable to benefit from information visualization due to the inherent complexity, scale, and real-time nature of network traffic. NetGrok fills this gap by providing visual analytics for real-time or archived packet captures using a grouped node-link visualization and a treemap to organize network hosts. Trusted local hosts are easily distinguished from external ones by an exponential polar layout, with convex hulls for known groups. Using dynamic filters, users can rapidly see only the time periods or subnets of interest. A case study showed that NetGrok serves as an ``excellent real-time diagnostic,'' enabling fast understanding of network traffic and easy problem detection. NetGrok was featured in a cover story on security visualization in Linux Magazine (PDF).

National Cancer Institute research network collaborations

Analyzed the network of collaborations between National Cancer Institute research organizations as part of a HCIL research group with Catherine Plaisant and Ben Shneiderman.

Picture of Cody Dunne

Email:

cdunne at cs.umd.edu

Address:

Department of Computer Science
Human-Computer Interaction Lab (HCIL)
2117 Hornbake Library, S. wing
University of Maryland
College Park, MD 20742