Vis Reading Group

The goal of our visualization reading group is to supplement each member's own research with group discussion and brainstorming. In the past, we have published group projects that formed out of these discussions.

In 2012-13, the reading group is called IVRG (InfoVis Reading Group). The theme will be Big Data visualization and analytics. We will primarily touch on papers from the InfoVis, VAST, and HCI communities, though scivis papers are welcome, too. The link to our public website and schedule is available here.

Paper Queue

Add title/author/link below.

• Hierarchical Aggregation for Information Visualization: Overview, Techniques, and Design Guidelines, Elmqvist et al., TVCG '10

• Incremental, Approximate Database Queries and Uncertainty for Exploratory Visualization, Danyel Fisher, LDAV '11

• Parallel Visualization on Large Clusters Using MapReduce, Vo et al., LDAV '11

• Now Where Was I? Psychologically-Triggered Bookmarking, Pan et al., CHI '11

• Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design, Heer and Bostock, CHI '10

• A Framework of Interaction Costs in Information Visualization, Lam, TVCG '08

• Big Data Storytelling through Interactive Maps, Madhavan et al., Bulletin of IEEE Tech. Comm. on Data Engineering '12

• Dynamic Reduction of Query Result Sets for Interactive Visualizaton, Battle et al., New England Database Summit '13

• Where People Look, Judd et al., ICCV '09

• Notes on Image Annotation, Barriuso and Torralba, arXiv '12

• Crowdsourced automatic zoom and scroll for video retargeting, Carlier et al., ACM Multimedia '10

• Design Principles for Visual Communications, Agrawala et al., CACM April '11

• CommentSpace: Structured Support for Collaborative Visual Analysis, Willett et al., CHI '11

• Perceptual Guidelines for Creating Rectangular Treemaps, Kong et al., Vis '10

• Sizing the Horizon: The Effects of Chart Size and Layering on the Graphical Perception of Time Series Visualizations, Heer et al., CHI '09

• behaviorism : A Framework for Dynamic Data Visualization, Forbes et al., InfoVis '10

• Human Computation in Visualization: Using Purpose Driven Games for Robust Evaluation of Visualization Algorithms, Ahmad et al., TVCG'12

• Scented Widgets: Improving Navigation Cues with Embedded Visualizations, Agrawala et al., TVCG'07

• Theoretical Analysis of Uncertainty Visualizations, Zuk and Carpendale, Proc. of SPIE-IS&T Electronic Imaging, 2006

Helpful Resources for Projects

• Programming HITs with the MTurk API, from Berkeley EECS's 2011 crowdsourcing seminar

Group Project Ideas

Extending the Slideshow InfoVis Paper

• We present work toward a new tool called pdf2ppt that uses a probabilistic generative model to create slideshow presentations from research papers. pdf2ppt generates the basic structure of a presentation so authors can spend less time on simple design tasks. We discuss the value of this tool in the context of information retargeting: redesigning information deliverables, like research articles, for different purposes or audiences. The generative model is trained using existing information about slideshow design conventions, which we found during an earlier ethnographic study. Finally, we describe findings from a user study that asked participants to make presentations with and without pdf2ppt. Participants found that producing slideshows with pdf2ppt was faster and made them more confident during the design process.--Steven Gomez 20:51, 10 September 2012 (EDT)

Surveying Visualization Use in a Domain

• We present a survey of visualization use and needs among graduate students in the biological and life sciences. The findings of our survey suggest a rough mapping between user-goals, like creating exploratory prototypes or production-quality graphics, and specific visualization methods and features. Commonly-used toolkits for visualization are evaluated with respect to their abilities to support these methods and features. Finally, we discuss holes in the design space for enduser-oriented visualization toolkits for the life sciences domain.--Steven Gomez 20:51, 10 September 2012 (EDT)

Crowdsourcing Interactive Visual Analysis for Brain Tractograms

• We present findings from an evaluation of interactive brain tractography analysis completed by online, crowdsourced workers. Crowdsourcing analysis can be used for evaluating visualizations without expert users, who can be difficult to recruit for traditional user studies. To the best of our knowledge, crowdsourcing has not been explored as a resource for completing interactive, scientific visualization tasks. Our goal is to validate the ability of workers to perform interactive analyses and contribute to crowdsourced evaluations, which can be executed quickly and with many users. We created a Web-based tractography visualization, defined a set of interactions and tasks, and recruited X workers from Amazon Mechanical Turk. We compared their results with the same tasks performed by a small set of brain experts. We found yadda ... yadda ... yadda.--Steven Gomez 15:03, 14 September 2012 (EDT)

Crowdsourcing Interactive Visual Analysis

• This is a more general version of the above. We present findings from an evaluation of crowdsourced interactive visualization tasks. (We systematically chose a set of tasks that embody a general set of unit interaction tasks.) These tasks were completed on (visualization X ... interactive versions of the Cleveland and McGill perception charts?).--Steven Gomez 15:03, 14 September 2012 (EDT)

Mining Online Visualizations

• We present a taxonomy of online visualizations created by publicly available visualization tools. We categorized x visualizations from [Many Eyes, name of other sources] by their subject area, data type and visualization type. We then discuss 1) the gap between people's data analysis needs and the availability of visualization techniques and 2) opportunities for visualization toolkits and/or methods that address the gap.--Hua Guo 11:51, 14 September 2012 (EDT)

Crowdsourcing Visual Saliency Maps in Large-scale Images

• The ability to predict salient parts of an image has many potential applications, ranging from human computer interaction to database design. Previous efforts have shown success with a data-driven approach where eye-tracking data and visual features are used to train a saliency classifier. In this work, we demonstrate a data-driven approach to learning saliency in large-scale images through crowdsourcing. We also explore task-dependence in salience models and how salience maps can be used for visual search tasks..--Ryan Cabeen