A Study On Number Of Issues Revolving The Speech Translation Between Mongolian And Japanese

Transliteration has been being active in theoretical and practical aspects at present.Speech translation of English,Chinese,Japanese,and other languages are used in phonological dictionary,tourism industry,artificial intelligence and some other fields.Speech translation of Mongolian is still under unformed.With the development of the society,the demanding needs to the translation between Mongolian and other languages are urged by the people.This paper makes a preliminary investigation on the speech translation system between Mongolian and Japanese through Python+ and Tensor Flow framework.This paper contains six parts which are introduction,the first chapter,the second chapter,the third chapter,the fourth chapter and conclusion.In the introduction,the reason for choosing the topic,research background,research content,research procedure and the research significance are introduced respectively.The first chapter illustrates the speech features and the approaches of extracting the speech features;at the meantime it introduces the acoustic model of Mongolian and the approach of constructing the speech model,decoding and so on.The second chapter introduces the deep neural networks of acoustic model in Mongolian.The third chapter a Mongolian speech recognition system based on Python + Tensor Flow framework is built and tested.The fourth chapter describes the approaches to implement the translation between Mongolian and Japanese based on the Seq2Seq+attention model and how to invoke the API methods of Japanese speech synthesis.In the conclusion,it concludes the whole study and clarifies the further work in the future.