Search result diversification enables the modern day search engines to construct a result list that consists of documents that are relevant to the user query and at the same time, diverse enough to meet the expectations of a diverse user population. However, all the queries received by a search engine may not benefit from diversification. Further, different types of queries may benefit from different diversification mechanisms. In this paper we present an analysis of logs of a commercial web search engine and study the web search queries for their diversification requirements. We analyze queries based on their click entropy and popularity and propose a query taxonomy based on their diversification requirements. We then carry out the task of automatically classifying web search queries into one of the classes of our proposed taxonomy. We utilize various query-based, click-based and reformulation-based features for the query classification task and achieve strong classification results.
All Science Journal Classification (ASJC) codes
- Information Systems
- Library and Information Sciences