Markov Model-Based Sentence-Level Input Method Algorithm Prototype Design And Implementation

This paper briefly introduces the Statistical Language Model, and discusses the design and implementation of pinyin-to-word conversion application by using Hidden Markov Model in detail. Surrounding the topic of how to efficiently implement Viterbi Beam Search, the paper discusses several topics, including how to compact disk utilization, accelerate the search algorithm by carefully handcrafting the data structure. Discussed the parameters of pinyin-to-word conversion, and provided how to get the parameters in this paper; design and redesign the storage format in the disk, maximize the performance to fit the need of sentence pinyin input method. At the end, we give out the correction rates by using the different smoothing methods.Succeed to modify the Hidden Markov Model to fit the need of pinyin-to-word conversion application, and the Viterbi Beam Search is based on word-unit, which is the most important part here.