Researching Of Long-Distance Mongolia Language Model Based On Trigger Pair

Language model(LM) is a mathematics model that describes inherent disciplines of natural language.It is occupying the extremely important status in machine translation(MT),word processing,information retrieve,speech recognition.A robust language model can greatly improve the result of MT.Mongolian is a worldwide influential national language and be widely used in many parts of China.Mongolian language model in the Mongolian language information processing systems integration and application technology has a certain development,but mainly uses Chinese,English and other languages, technology,and has not addressed the Mongolian itself a substantial amount of research.With the Mongolian language model technical aspects of the research results,this paper will explore further Mongolian,through enhanced Mongolian language model to improve the performance of the Chinese-Mongolian Machine Translation System recognition rate.In this paper,the relevant technical language model,the characteristics of the Mongolian language,trigger features were discussed;the long-distance Mongolia language model based with the characteristics of the Mongolian language and trigger features was designed;smoothing techniques was discussed;with the trigger features,relevant KATZ smoothing algorithm was designed;the experiment result indicated that the long-distance Mongolia language model based on trigger pair had better performance and satisfactory results of the translation.In the case of translation nodes being 300,the bleu score rises from 0.1761 to 0.2234;in the case of translation nodes being 1000, the bleu score rises from 0.1873 to 0.2325.