An Extended Approach to Personalize the Web Search by Measuring the User Relevance

Abstract

Current methodologies in personalizing the web search include number of hits on a link by the user, inbound links and user profiling which are independent of user’s action. In this paper we present a new dimension of attributes which can be used to get more personalized search results. These attributes are based on the user actions after he/she has opened the given link. The attributes include actions like saving/printing/bookmarking the document etc. which are clear indicators of user relevance. Simulation and experimentation of the same using a web search interface and comparison of search results using precision and recall graphs showed an average of 22% improvement in the relevance of the search results returned, ranging from 0-60% for different weighting models.