Sentence completion using Text Prediction systems

Abstract

Text Prediction for sentence completion is a widely used method to enhance the speed of communication as well as reducing the total time taken to compose text. This paper briefly describes the approaches, design and implementation issues involved, as well as the factors and parameters that determine effectiveness of a system. The information is then used to build a software system, capable of modeling text data, in order to generate predictions in real-time. By using a pure statistical approach, we generate N-gram models that are adaptive to users by applying instance based learning. Details of the software development method, used to prototype and iteratively build a highly effective system, are provided.