Evaluating Visualization Techniques Using Eye Tracking
After 5 years of working on various
visualization projects, I've decided that instead of inventing new
visualization techniques, it's more important to understand better
why the techniques we have work (and why they don't). My work
in this area is focused
on measuring the task-evoked pupillary response as a way of
estimating the cognitive load required by visualization tasks.
I've demonstrated that I can use a remote eye tracker to measure
changes in pupil size caused by cognitive load. I published this
result as a paper
in ETRA 2008. My stimuli for that first study were
all audio, and I'm moving on now to measurements
of cognitive load changes during the completion of visual tasks such
as visual search, mental rotation, and visual comparison.
past projects
Gaze-Enhanced User Interface Design
In the spring of 2007, I did a bit of work
with Manu Kumar on his GUIDe project,
in which we explored real-time smoothing
and timing-correction algorithms for gaze data in order to reduce input errors
and make gaze more viable as a practical input method. We published these
techniques in a a paper
in ETRA 2008.
Visualizing Heterogeneous Data
At the beginning of 2007, I joined Mike Cammarano and Bryan Chan on
this project exploring the automatic visualization of semi-structured data.
We published a paper on this work in
InfoVis 2007.
Ranking for Graph Drawing
How can ranking algorithms help in
the clean layout of graphs? When a graph is so dense you can only
draw a small piece of it, which subgraph should you draw? In Fall 2006,
I developed
a flow-based algorithm similar to PageRank to select a relevant subgraph
based on a focal node and user-supplied hints about which types of
nodes and relationships are important.
Graph Tables
I spent the 2005-2006 academic year working on
a formal algebra of graphs implemented on top of relational algebra.
This algebra is aimed at making it very easy to extract and flexibly visualize
the various graphs present in a relational database. The best way to see
what this is about is to watch the video demo.
Congressional
VoteViewer
The US House of
Representatives Clerk's
office publishes the records
of all roll call votes taken in Congress since 1990. This data
tells
you how
every congressperson voted on every bill and procedural issue. It's
great data, but it
doesn't come in a very usable form. In the spring of 2005, Mike Green and I made a web-based
visualization tool to make
this data more accessible. It's based on matrix permutations and is
focused on revealing
large-scale voting patterns in congress. When I get some free time, I'll
post the web app here. Stay tuned.
Visual Exploration
of Citation Networks
In early 2004, I crawled
the ACM digital library
and grabbed info and citation links for all SIGGRAPH papers from 1974-2003. I
made a graph-based
visual exploration tool for exploring the
collaboration and citation networks of SIGGRAPH papers and authors.
This project was
a preliminary exploration of some of the ideas I'm exploring in my
thesis research
on using graph drawing to explore general relational data.
Automatic View
Selection
Can a good viewing angle
for a 3D object be chosen automatically? What
makes a viewpoint good (or bad)? I spent some time applying
the mathematics of information theory to this problem during 2003.
Since I started up with the graph drawing stuff, this research
has gone to the back burner.
Image-Based
Relighting for Illustration
Photographs can be very
useful for documenting an obejct's appearance,
but the lighting control needed for a good photo can be difficult or
impossible to achieve. Dave Akers, Frank Losasso, and I made an
image-based relighting system that can be used to make photographic
composites that effectively convey an object's shape and features. It
was inspired by the techniques of lighting design and illustration. We
presented this work in a paper
at the Visualization
2003 conference in Seattle in October 2003. Michael Cohen and some
other folks at MSR and UW have done
some
really cool releated work that goes much farther than we did. They
showed their "Interactive
Digital Photomontage" paper at siggraph 2004.
Generating
Step-by-Step Assembly Instructions
Everybody knows how
frustrating most assembly instructions can be.
During the fall of 2002, I worked with Maneesh Agrawala, Doantam Phan,
and Pat Hanrahan in C.S and Julie Tversky from the psych department on
a software system that generates effective assembly instructions
automatically. We presented this work in a
paper at siggraph 2003.
Phylogenetic Tree
Set Visualization
I worked with some great
folks at the University of Texas and the City University of New York on
the problem of visualizing
lots and lots of evolutionary trees. We came up with some neat
software
that the biologists there liked. A write-up of this work constituted my
senior thesis at UT Austin. This project is still going strong; you can
read about recent developments at its new website.
