Lighthouse originally was developed as a document clustering and visualization system. It combines traditional ranked list metaphor with clustering and spring-embedding visualizations. It also includes an adaptive relevance feedback component based on neural network.

Lighthouse has evolved into a information access platform that includes a search engine. It has been successfully integrated with other language technologies such as multi-document summarization, machine translation, cross-language retrieval, headline generation, etc.

This page contains links and information about the Lighthouse system.

• The Lighthouse demo. The demo can be used as either applet or application.
  
  News [August 2003]. The Lighthouse demo is presently off-line. The server machine has been retired by the CIIR system support. I am working on building a platform neutral version of the server. I estimate it to be completed in September.
• A screenshot.
• A movie of my presentation at InfoVis 2000 that demonstrates some of the features in Lighthouse
   
  • Large: 640 x 480
  • Small: 320 x 240
• Lighthouse has wrappers for different web search engines. For example, a user can post a query into the interface, receive the search results from Google, visualize and browse them in Lighthouse. Lighthouse downloads and parses the web pages automatically to construct the visualization.
• We extended Lighthouse to work with streamed news stories data. Lighthouse visualizes clusters of news stories, dynamically updating the picture as new stories arrive into the system. A user can get a good sense of the new stories from their spatial proximity to the existing clusters.
• I have integrated Lighthouse with an open-source search engine Lucene. A user can index and search files on the local hard drive.
• I have extended Lighthouse to handle email data. A user can search and navigate email messages and people records.
• I have integrated Lighthouse with the multi-document summarization system, headline generation system, and also added query translation capabilities, so it can be used for cross-language retrieval.