DiscoWeb: Discovering Web Communities Via Link Analysis

Department of Computer Science
Rutgers University

The DiscoWeb group investigated the use of link analysis on the WWW to help rank search engine retrievals and to help generate web directories. By examining the graph structure of the web, our system is able to automatically select highly-regarded web sites out of the many thousands of pages that are considered relevant to a query by a search engine. The name of our group comes from one of the capabilities of our system: it can discover web communities -- highly interconnected web sites that typically share a single topic or focus.

In May 1999 we presented a poster at the Eighth International World Wide Web Conference in Toronto. The extended abstract from that poster is available.

We have two prototypes that incorporate this link analysis. One of them is a general purpose search engine and we plan to make this engine publicly available as soon as it is robust enough for public use. An overview of the architecture of this system is available.

Last modified: 16 July 1999 by Brian D. Davison.