In each case, I've listed the general area but not really the task that a user would be trying to accomplish in that area. It is important to identify the target task(s) before designing the visualization. It is also important to think and describe how the visualization would fit within the context of a user interface that helps users work on their problems as a whole.
In the WebTango research project, we are developing tools to help web site designers evaluate sites, both to consider alternative designs and assess what is problematic about current designs.
There are two visualization problems in this domain.
One project is to design a visual interface for displaying the results
of a metrics analysis tool we are developing to help users improve
their site designs. The tool does not yet do a site-level analysis,
but we would like it to in future. To discuss ideas realted to this,
send mail to
Another project is to help assess the usability of the site via examining
usage data in the web logs. There are well-known problems assiociated
with doing this, but it should be useful as well. The implementation
of this project should use existing data since it is time-consuming to
collect new data. There is some available at the KDD Cup 2000 site, as well
as a discussion of the analysis of this data using data mining
techniques. This problem is tough because there is a huge amount of
data, it has complex structure across time and many different users,
and it is incomplete.
Some students (Tim Hirzel, Andrew Volpe, and jeff Enos). made an interesting attempt last year.
Their writeup.
Figure 1.
Figure 2.
For other examples, see:
The QUIP
project (not applied to the web, but an interesting related idea).
Gerald L. Lohse, University of Pennsylvania; Peter Spiller, McKinsey &
Company
Quantifying the Effect of User Interface Design Features
on Cyberstore Traffic and Sales Proceedings of ACM CHI 98
Conference on Human Factors in Computing Systems, 1998. (This is a nice
empirical evaluation of 20 online (web-based) "superstores" comparing
how various aspects of their web site design and information layouts
correlate with sales and traffic. Regression analysis was used to
assess impact of variables.)
A different but related problem is visualizing an individual's web usage history
in the hopes of helping them navigate:
Michael D. Byrne and Bonnie E. John and Neil S. Wehrle and David
C. Crow,
The Tangled Web We Wove: A Taxonomy of {WWW} Use
Proceedings of ACM CHI 99 Conference on Human Factors in Computing
Systems, Volume 1, 544--551, 1999.
Tauscher, L. and Greenberg, S. (1997).
How People Revisit Web Pages: Empirical Findings and Implications for
the Design of History Systems. International Journal of Human
Computer Studies, Special issue on World Wide Web Usability, 47(1),
p97-138. Academic Press.
We have
the problem descriptions, sample data, and her talk.
The goal of this class project would be to think of new kinds of
visualizations that work well within this paradigm for particular
types of datasets. For example, Hellerstein et. al have given
examples of progressively showing university enrollment data by
plotting the data progressively as an average with error bars that
shrink as more data becomes available. They've also shown cities on a
map, showing "clouds" of points, starting with the densest parts of
the population, rather than alphabetical order by state, thus showing
the most important information first. How else can visualization of
information from large datasets be done differently given the
underlying mechanisms supplied by the system?
Project groups should have 2-3 people.
Timing:
Grading:
Harry Hochheiser, Ben Shneiderman
Using Interactive Visualizations of WWW Log Data to Characterize
Access Patterns and Inform Site Design
Ron R. Hightower et al.,
Graphical Multiscale Web Histories: A Study of PadPrints
Visualizing Database Systems and Content
Anna Wichansky
of Oracle described a number of projects relating to
database systems. These include
Incrementally Revealing Database Content
The CONTROL Project, led
by Prof. Joe Hellerstein at UC
Berkeley, has the goal making queries against large databases more
interactive. This is done by statistically sampling from the contents
of the datbase, and showing the results to the users visually. As
more data is retrieved, the visualization becomes more complete.
(CONTROL stands for Continuous Output and Navigation Technology with
Refinement On-Line), here is an easy-to-read
overview.
Add Features to XMDVtool
Add features to
Matt Ward's visualization tool,
XmdvTool.
Some specific suggestions include:
Evaluate Efficacy of a Visualization
Perform a formal evaluation that compares a visualization
technique for a particular task to either another visualization
or a non-visual technique. (See http://otal.umd.edu/SHORE
for examples of user interface evaluations performed by students
in Ben Shneiderman's HCI course.)
Project Process
Submit Project Proposal Here
