User dictionary merge for enhancing smart phone auto prediction

Abstract

Smart phone keyboards render a device based text prediction system which is usually built around the keyboard application under use. Withal, they infer a restriction since the techniques used in the system are mostly confined to device in scope. In this paper, we formulate a method to merge various user dictionaries so as to increase keystroke savings on average for a smart phone user. The method described in this paper takes into account relations among a multitude of users, and applies it to merge the user dictionary at a level derived from the relations. We show an increase in percentage keystroke savings for experimental data by applying this algorithm to context insensitive variation of word prediction.