Navigation

• Main Page
• About

• Random Page
• Create a new Page
• All Pages
• Categories

• Administration
• File Management

• Login/Logout
• Language Selection
• Your Profile

Quick Search

Advanced Search »

Graph-based methods for visualizing knowledge are a powerful means of conveying information. Of particular interest are visualization methods that are interactive and animated.

TableOfContents [HideShow] The Visual System Self-Organizing Maps     WebSOM     ThinkMap Other Graph Layout Approaches     GraphViz     TheBrain     TouchGraph    Jambalaya Applications of Graph Visualization    Visualizing the Semantic Web    Visualizing Belief Networks    Visualizing the English Language Links

Edit

The Visual System

Human-computer interactions involving complex knowledge representations are potentially much more efficient with this type of user interface, because the use of spatial relationships and interactivity is a better match for the full capabilities of the human visual system.

There are two main processing streams within the visual system:

• the Temporal stream processes symbolic information and shapes
• the Parietal stream processes spatial relationships and changes in those relationships (corresponding to egocentric motion, for instance)

Most abstract information is presented in a format intended for Temporal stream processing (i.e. in a symbolic format). Part of the rationale for graph-based visualization is that, by utilizing the spatial processing stream, visualization can more effectively utilize the entire processing capacity of the visual system.

Edit

Self-Organizing Maps

The approach that we have used, and that used by some of the other graph visualization systems, is based on work done by the mathematician Tuevo Kohonen on self-organizing maps, which are a type of unsupervised learning algorithm. We have extended the original SOM to allow for dynamic, incremental updates. The Dynamic SOM is the basis of the more powerful layout algorithm within theWheel.

Our implementation also has a faster, real-time mapping called Dynamic Sammon's Mapping. This is used in conjunction with the Dynamic SOM to quickly layout nodes as the user interaction changes the activation of nodes of interest.

The self-organizing map approach has been applied to a number of research and commercial applications.

Edit

WebSOM

An application of the original SOM formulation to document management. Done by Kohonen's group in Helsinki , provides a very interesting navigational mechanism for web documents, patent collections, and other other large collection of documents.

Edit

ThinkMap

A nice web-based graph visualization tool that appears to use a pseudo-SOM to lay out nodes consisting of single words or phrases. Interesting animation effects (see the Visual Thesaurus in particular).

Edit

Other Graph Layout Approaches

Other graph visualization tools use various techniques to spatially arrange information.

Edit

GraphViz

From BellLabs is an open source graph layout engine. It includes node layout as well as routing of edges. The node layout is fairly simplistic (mostly just ordered arrays) but the edge layout is fairly intelligient.

Edit

TheBrain

Uses an algorithmic layout engine to arrange groups of nodes. Not as imaginative as the Star Tree or Thinkmap, but it seems to handle graphs with a larger average fan-out better than the SOM-based approaches.

Edit

TouchGraph

has a library and set of applications that use force-directed graph layout.  Very dynamic and responsive.

Edit

Jambalaya

is a plug-in for the Protégé Ontology Editor that draws very nice graphs of knowledge structures.

Edit

Applications of Graph Visualization

Edit

Visualizing the Semantic Web

Graph visualization is becoming increasingly important as a user interface because of the close connection between graph representations and more sophisticated knowledge representation methods. For instance, the Semantic Web, which is being defined now by the W3 Consortium, is a generic means of representing information about the Internet using directed graphs.

Edit

Visualizing Belief Networks

Directed graphs are also an effective means of representing probabilistic inference in the form Bayesian belief networks. The visualization and navigation of large belief networks is greatly enhanced by sophisticated graph visualization. Judea Pearl at UCLA has done some interesting work on the representation and manipulation of Bayesian belief networks. A new company, Numenta, has been started to develop commercial applications of Pearl's approach.

Edit

Visualizing the English Language

The Visual Thesaurus and Visuwords are interesting applications of animated graph displays of the English language, with links derived from meanings. The links for Visuwords are taken from the WordNet ontology.

Edit

Links

• Information Visualization on Open Directory
• The InfoViz:Wiki