The Research On Constructing Layered Mongolian Laguage Model

Language Model (LM) is a mathematics model that describes inherent disciplines of natural language.It is applied in the field of natural language processing that contains Machine Translation, Speech Recognition and Text Collation. Nowadays, the construction of Statistical Language Model is becoming a new hotpoint to research, and the Statistical Language Model is becoming a prevailing technology of natural language processing gradually. At present, a lack of researching on Mongolian Language Model has restricted the course of informationization.This paper has discussed the correlative technic of Statistical Language Model, and has analyzed the characteristics of Mongolian. According to the characteristics of Mongolian, try to improve the veracity of Language Model by reasearching dependent relation between Etyma and Affix. The method is dividing Language Model into three administrative levels, they are Etyma and Etyma, Etyma and Affix, Affix and Affix. Construct Language Model on these administrative levels separately, and then band them together. So this Language Model is adapt to Mongolian.This Language Model has been applied on a Chinese-Mongolian machine translation system that is based on example. The experiment proves that the Language Model could improve the translation result.