Intent-Based Categorization of Search Results Using Questions from Web Q&A Corpus

Abstract

User intent is defined as a user’s information need. Detecting intent in Web search helps users to obtain relevant content, thus improving their satisfaction. We propose a novel approach to instantiating intent by using adaptive categorization producing predicted intent probabilities. For this, we attempt to detect factors by which intent is formed, called intent features, by using a Web Q&A corpus. Our approach was motivated by the observation that questions related to queries are effective for finding intent features. We extract set of categories and their intent features automatically by analyzing questions within Web Q&A corpus, and categorize search results using these features. The advantages of our intent-based categorization are twofold, (1) presenting the most probable intent categories to help users clarify and choose starting points for Web searches, and (2) adapting sets of intent categories for each query. Experimental results show that distilled intent features can efficiently describe intent categories, and search results can be efficiently categorized without any human supervision.