The Research On The Mongolian朎 Nglish Machine Translation System Based On A Hybrid Method

Machine translation (MT) is defined that the text is translated from a language to another language by computer. With the widely communicating with outer world, machine translation from some language to English is bestowed on the great practical significance. Meanwhile, the research on machine translation will also put forward the study on computational linguistics, natural language processing and artificial intelligence. Moreover, the study of machine translation is bestowed a very important significance on the collection of information.The methods of machine translation can be classified as rule-based method and corpus-based method. The language characteristic rules can be accurately described by the rule-based method, but not all language phenomenons could be covered. The corpus-based method can be divided into the example and the statistical method. No matter the example method or the statistical method all need the massive corpus. In this thesis, a hybrid machine translation method, which will be mainly based on the example method and the auxiliary method: the rule and the statistical method, is proposed.We implement a Mongolian-English MT system, which is composed by three modules. The first module is designed to automatically acquire the translation fragment pairs, in which module the statistical and the hypothesis-testing method are used. The second module is the translation module based on the example method. In this module, we mainly consider how to search, match and analogy. The sentence which can't be translated by the second module will be translated in the last module, adopting the similar "break-section" translation method, which can be called the hybrid rule and example-based method in this system.Finally we do some testing, summarize the strongpoint of this Mongolian-English MT system, and discuss the prospect of MT.