Information Visualization and Visual Analytics
Team Members: Mengdie Hu, Hannah Pileggi, Zach Pousman, Ramik Sadana, Chad Stolper, John Stasko
 |
|
|
The research areas of Information Visualization (InfoVis) and Visual Analytics (VA) are two of the main foci of the II Lab. This work is all about presenting data visually in order to aid people in exploring, analyzing, and understanding the data. Information visualization typically focuses on abstract data, that is, data without any agreed-upon depiction, such as financial data, text, statistics, databases, and software. Visual analytics emphasizes analytical reasoning about data and combines computational analysis techniques with interactive visualizations. Below are a number of the subareas within InfoVis and VA in which we are working. For a broad look at InfoVis/VA related links and information on the web, see our InfoVis web resources page.
Theoretical Foundations - We are exploring the application
of distributed cognition as a theoretical foundation for
information visualization. The concepts inherent in distributed
cognition appear to be a very sound basis to understand the value
of infovis and how to make visualizatyions more effective. We
also seek to better understand the value of visualization as a
cognitive aid.
Techniques and Tools - Our work often involves the
development of new visualization techniques and tools. Sometimes the
techniques/tools are built for specific kinds of data, and other
times they are more general purpose. We have developed techniques
to portray tabular data, video data, and hierarchical data, among
others, and systems such as Ploceus, SellTrend, and Dust 'n Magnet.
Domain-specific problems - We frequently work with people
and organizations from different areas and who have data that they
want to understand better. Recent examples include work in areas
such as web transactions, airline travel, business markets, mutual
funds and stocks, RFID location information, and sports. We
build interactive visualization systems that help these people
explore and analyze their data.
Casual InfoVis - Traditional or "core" infovis typically
refers to "deep-dive" analysis often done with sophisticated,
multiple-view systems used for hours at a time to do complex
analysis. Another style of infovis that we call Casual InfoVis,
still involving the visualization of information, involves the use
of visualizations for just a few moments at a time but very
frequently. We are defining and exploring this new subarea of InfoVis.
Interaction - InfoVis is made up of two main components:
representation and interaction. Representation gets much more
focus and is often thought of as more exciting. Interaction, the
"little brother", does not receive as much focus, but we argue
that it is crucial for InfoVis and harbors the potential for
greater innovation in the future.
User Tasks - Too often InfoVis is about innovative visual
representations with little concern on whether those
representations are actually useful to people. We have been
promoting a more task- and
analysis-centric view of the discipline, rather than a
representation-centric view. In a paper at InfoVis '05, we
identified a set of 10 low-level analytic tasks such as
correlate, filter, cluster, etc., that
characterize what people do when using infovis systems in data
analysis. In papers in TVCG '05 and InfoVis
'04, we posited a set of higher-level knowledge precepts that
help bridge the analytic gaps between data representation
and higher-level analytic tasks, such as forecasting, learning new
domains, and cost-benefit analysis. For more details, see the vis tasks webpages.
Evaluation - Evaluation has been a key theme of our
research and we make sure to stress careful evaluation in all of
the infovis projects from the lab. This topic is certainly
related to our work on User Tasks, since clearly understanding
what a person is trying to accomplish with a system is key to
evaluating the system well. More recent work has focused on the
ICE-T visualization value evaluation approach.
Hierarchical Data - We have done a lot of work on
visualizing hierarchies through different kinds of space-filling
visualizations. In 2000, we introduced the SunBurst circular space-filling
technique/system that uses animation to help drill down into the
hierarchy and we compared SunBurst to the well-known Treemap
technique. At InfoVis '03, we also introduced a variant of
Treemaps called the Context Treemap
that has been useful for portraying data sets like mutual fund
portfolios. Software Visualization - This is a large subarea of Information Visualization that involves visualization of software, both algorithms and programs, to help people better understand the software. We have been researching software visualization for twenty+ years, and our work is summarized in the SoftVis web pages. The softvis.org web pages provide further information about the area including links to the ACM Symposium on Software Visualization.
Information Security -
Together with information security researchers at
Georgia Tech, we have examined how visualization may
assist information assurance and information security. We've
developed systems for visualizing IDS alarm and network packets.
Education - John Stasko has been teaching
CS
7450, a graduate course on Information Visualization, since 1999.
His syllabus, lecture slides, assignments, and bibliographic
entries are all provided on-line at the class website. Video
recordings of almost all the class lectures from the course also
are available in the Visual
Analytics Digital Library. Conferences - John was General Chair for IEEE VIS 2013 in Atlanta. Previously, he was Papers Co-Chair for the IEEE VAST 2009 Symposium. He was General Chair for IEEE InfoVis 2007 and Papers Co-Chair for InfoVis 2005 and InfoVis 2006. Some subset of Information Interfaces students always attend the conferences and our group is an active paticipant year in-year out.
Past and Present InfoVis and Visual Analytics Systems Research
Projects Papers |