theWheel navigating a music collection
theWheel is a tool for knowledge management and
knowledge visualization. It works with knowledge represented as interconnected pieces of information, known as graph-based models. This is the knowledge representation approach followed by the
Semantic Web, where knowledge is encoded as a graph consisting of nodes (representing resources), connected by annotated links (representing relationships among the resources).
Knowledge bases that can be represented this way include:
- A reaction pathway map for a signal transduction network
- A neural connectivity map
- A set of cross-referenced scientific papers
- A music collection
- A family tree
theWheel manages the nodes in the knowledge base by tracking the links connecting the nodes. The link between any two nodes is assigned a weight, so that two nodes (two papers, two songs, two recipes) that are closely related have a high link weight. If two nodes are not at all related, they have a link weight of zero.
A group of nodes that are interconnected form a space. Interactive visualization of the space using the theWheel is intuitive. One of the nodes in the space is designated as the entry point into the space. When the space is first opened in theWheel, this node is displayed prominently near the center of the main window into the space (the space view). Other closely related nodes are displayed around it. To traverse the space, the user only needs to move the mouse to indicate interest in a particular node. As the user's interest in the nodes shift, the display is dynamically updated.
An animated
screen capture shows theWheel in action. theWheel can be downloaded
here, and you can also read the
online help.
EditHow does it work?
theWheel uses an internal knowlege representation consisting of nodes connected by directed, weighted links, where the link weight between two nodes is proportional to the relatedness of the nodes.
theWheel tracks user interest in the different nodes by calculating an activation for each node. The activation value varies smoothly between zero and one. As the user traverses the space by moving the mouse over nodes of interest, this causes those node's activation to be increased proportional to the amount of mouse motion (this is called wave activation, or primary activation). The increase in the primary activation of the nodes triggers a propagation of the activation to related nodes (this is called propagated activation, or secondary activation). The amount of propagated activation of the related node is proportional to the weight of the link connecting the nodes.
This update of activation values occurs many times per second. After each activation update, theWheel calculates a new layout for the nodes so that nodes with higher activations are given more space, and nodes that are related are placed closer together. The algorithm that does this is designed so that the incremental changes in the layout resulting from primary and secondary activation will tend to be very smooth from moment to moment, creating a compelling animated visualization of the space.