Inferential language modeling for selective Web search personalization and contextualization

Abstract

Personalized Web search systems have been explored to alleviate the problem of information overload by keeping track of a user's specific information retrieval (IR) preferences, and then pushing information to the user according to their preferences maintained in a user profile. Nevertheless, personalization and contextualization is always associated with a computational cost. Therefore, it is more advantageous for a personalized Web search system to evaluate the necessity of personalization for a query before invoking the personalization mechanism. Unfortunately, most of the existing personalized Web search approaches only blindly personalize users' queries without considering the characteristic of the queries or the searchers who issue those queries. The main contributions of our research work presented in this paper are two fold. First, a novel selective Web search personalization and contextualization method is developed to enhance the effectiveness of personalized Web search. Second, an inferential language model which can take into account the semantic and contextual information associated with a Web search scenario is developed to enhance the selective personalization and contextualization process. The results of our initial experiment show that the proposed selective personalization and contextualization method underpinned by inferential language modeling significantly outperforms a baseline method developed based on syntactic click entropy. To the best of our knowledge, this is the first inferential language modeling approach that has been successfully applied to Web search personalization and contextualization.