Mining User's Location Intention from Mobile Search Log

Yifan Sun, Xin Li, Lin Li, Qi Liu, Enhong Chen, Haiping Ma

Abstract: Much attention has been paid to web search personalization and query optimization over the past decade. With the prevalence of smart phones, the mobile search results for the same query may vary in regard to the user’s location. In order to provide more precise results for users, it’s essential to take geographic location into account along with the user’s input query. In this paper, we try to identify queries that have location intentions. For example, query “weather forecast” has a location intention of local city while “The Statue of Liberty” has a location intention of “New York city”. To identify the location intention behind a query, we propose a novel method to extract a set of features and use neural network to classify queries. In the classification of queries without explicit location names, our experiment shows that our approach achieves 82.5% at F1 measure and outperforms baselines by 4.2%.