Automatic generation and use of negative terms to evaluate topic-related Web pages

Abstract

Deciding the relevance of Web pages to a query or a topic is very important in serving Web users. For clustering and classifying Web pages the similar decisions need to be made. Most of work usually uses positively related terms in one form or another. Once a topic is given or focused, we suggest using negative terms to the topic for the relevance decision. A method to generate negative terms automatically by using DMOZ, Google and WordNet, is discussed, and formulas to decide the relevance using the negative terms are also given in this paper. Experiments convince us of the usefulness of the negative terms against the topic. This work also helps to solve the polysemy problem. Since generating negative terms to any topic is automatic, this work may help many studies for the service improvement in the Web.