End-To-End Algorithm Optimization and Implementation Based on Speech Translation

Abstract

Speech translation is the process of automatically translating the commands from the speech signals of the source language into the text of the specified language by using machine completion instructions. The main operation method now is to associate speech recognition with machine translation. Because speech recognition, machine translation and speech translation are essentially the transformation from one sequence to another, and the end-to-end architecture can also be applied to speech recognition and machine translation, the researchers studied end-to-end speech translation. Because the integration of speech recognition and machine translation into one model modeling makes the input sequence and output sequence more difficult and cumbersome, it is necessary to carry out more difficult training on the model and analyze more training data. On this basis, this paper studies end-to-end speech translation and related training methods respectively.