Corpus-based predictive text input

We introduce a new predictive text input system that uses visited documents for predicting the user's next input word. With our method, users can efficiently compose new texts using other documents as dictionaries for the input prediction. We have developed a document storage system Kukura that stores all the texts visited by users and cooperate with our predictive input system PRIME. In this paper, we describe the design, implementation, and evaluation of our approach.