Using context to improve query formulation and entry from mobile phones

The goal of this thesis is to improve the query formulation and entry step of the web search process from mobile phones. We achieve this by providing query recommendations (queries displayed to the user before she has begun query input) and query predictions (queries displayed to the user as she is entering the query) as part of the mobile search interface.;To address this, we present two approaches for improving mobile query formulation and entry: relevant query recommendations and accurate query predictions. We build recommendation and prediction models that consider a user's context when she is interacting with the search interface. We restrict context to the set of circumstances captured in the Google search logs at the time of query. We find that the two contextual signals that make the biggest impact in improving the accuracy of predictions, thus reducing the number of key presses needed to enter a query, are knowledge of the application being used (in this case a search engine), and of the location of the user. Although knowledge of the day-of-week did not significantly impact the prediction model, it improves the quality of the recommendations. Knowledge of the time-of-day did not significantly improve either model.;The query recommendation and prediction models presented in this thesis are general and can be used in conjunction with any search interface. However, we focus our research on search interfaces for mobile phones, where we believe the greatest need for improvement exists. We design, implement, and evaluate interfaces for the prediction and recommendation models in order to improve the mobile search experience. Our findings can be easily incorporated in existing mobile search systems.;We derived the motivation for our research from a comprehensive overview of the state of mobile web search. After analyzing millions of requests made to the Google mobile search interface, we were able to identify areas for improvement. We discovered that the first step of the search process---formulating and entering the query---is time-consuming and cumbersome for mobile users.