Abstract: Overherd is an information visualization tool that makes patterns of interaction and activity in  online discussion boards visible so that instructors can more easily analyze their students’ contributions. Our system uses the forums tool from the Sakai learning management system (LMS) to create visual representations of online behavior. We describe the design and development of Overherd, and provide design recommendations for “mashups” that extend the functionality of LMS yet rely on existing data within these learning environments.
This work is done with Dr. Stephanie Teasley and Dr. Libby Hemphil at U of M.  See for more information and the code.

The following briefly describes a working prototype I built.

The tree map view of the discussion thread. Each blue-ish rectangle represents a leaf node (the last post) in the tree structure.  The intensity of the color represents the depth and posts can be filtered out by their depths.

The right dark panel shows the content of each post when you click a blue node.  You can also search for keywords and filter out nodes by date using the control at the bottom (as shown below).  A graph shows the number of new posts.

Overherd also visualizes the conversation network among the students.  In the below screen, each circular node represents a student and each link indicates there was at least one conversation among students.  When clicking a neighbor (a student with whom the currently selected student had a conversation), the conversations between the students are shown in the left panel.  A rudimentary statistical text analysis is also performed to find keywords in the conversations, which are highlighted.  I also wrote a parser in Ruby that can take excel data and create a TreeML format used by Overherd.