The Research On The Example-based Mongolian And Chinese Machine Translation

Machine translation refers that a computer automatically translates one type of natural language into other type of natural language which is needed by users. Machine translation, as a branch of natural language information processing, has experienced a long and bumpy course of development. It has walked through resplendence, dire straits, recovery. In recent years, with the improvement of computer hardware, the example-based machine translation(EBMT) has been the subject of much speculation. EBMT has overcome the difficulty of obtaining knowledge when using rule-based machine translation(RBMT). It used existing instance database information instead of tedious linguistic analysis, on the basis, it used similarity algorithms to match the most similar instances and reconstructs the sentences. After those, final translation were given. This method has the advantages of accurate and faster.This paper firstly discussed the purpose and significance of machine translation research, it introduced the present state of the machine translation research, especially expounded the Mongolian information processing and Mongolian to Chinese machine translation. Secondly, this paper discussed the development history and varies methods of machine translation, especially generalized basic principles and system architecture of EBMT methods. Thirdly, this paper described the construction method of Mongolian and Chinese bilingual instance library, it reached different particle size(paragraph level, sentence level, word level) alignment technology which is used in processing of pretreatment of bilingual corpus. Then, this paper discussed and arrived at sentence similarity algorithm based on words and the most similar example retrieval algorithm.In the end, this paper carried out the experiment of the example-based Mongolian and Chinese machine translation. Besides, this paper analyzed the result through the evaluation system. After results analysis, the paper summed up the advantages and disadvantages of the method. Experimental results showed that the translation achieved the expected goal that it could translation simple Mongolian sentences into understandable Chinese, particularly, if there is a match between the sentence needed to be translated and instance which was existed in the instance library, the system could get higher quality translation.