An Algorithm for Constructing Higher Order N-grams of Chinese Words

摘要

N-gram statistical language model is a fundamental component of a speech recognizer and is widely used for improving the performance of handwriting recognition, Chinese keystroke input, and many other apllications. N-gram with larger number of N(higher N-gram) is supposed to perform better than small ones. In this paper, we propose a new higher N-gram construction algorithm. We show that the algorithm is efficient than time-consuming old method and has the advantage of inheriting the existing language model.