On Semi-Automated Web Taxonomy Construction

The subject of this paper is the semi-automatic construction of taxonomies over the Web. We address the problem of discovering high-quality resources that belong in a particular node of a taxonomy. We show that minimal additional effort is required to provide relevance feedback in a hyperlinked environment, resulting in significant and consistent improvement in quality. Furthermore, this feedback is especially valuable for topics for which it is more difficult to find high-quality pages. Enroute, we describe novel algorithms for hyperlink relevance feedback.

• Type: Conference
• Year: 2001
• Where: WEBDB
• Authors: Ravi Kumar, Prabhakar Raghavan, Sridhar Rajagopalan, Andrew Tomkins
• Keywords: Database, Hyperlink Relevance Feedback, Minimal Additional Effort, Relevance Feedback, WEBDB 2001