Translation model-based training and translation methods and apparatuses, and storage medium

Abstract

Embodiments of the invention disclose translation model-based training and translation methods and apparatuses, and a storage medium. The training method comprises the steps of inputting a source statement to a translation model, thereby obtaining a target statement output by the translation model; determining the fidelity of the target statement to the source statement; by taking the target statement and a reference statement as inputs of a judger model and taking the fidelity as an output of the judger model, training the performance of calculating the similarity between the target statementand the reference statement by the judger model; outputting the similarity between the target statement and the reference statement by using the judger model; and by taking the source statement as aninput of the translation model, taking the target statement as an output of the translation model and taking the similarity as a weight coefficient, training the performance of outputting the corresponding target statement according to the input source statement by the translation model.

Description

technical field [0001] The present invention relates to machine translation (MT, Machine Translation) technology, in particular to a translation model-based training method, translation method, device and storage medium. Background technique [0002] With the development of machine translation, Neural Machine Translation (NMT, Neural Machine Translation) has been widely used as a new generation of translation technology. The neural network machine translation system is built based on the encoder-decoder framework. However, in the translation process of the neural network machine translation system, the decoder has multiple tasks, such as recording the current translated content and the content to be translated, and recording the translated content. Fluency related information, etc. Due to the complexity of the decoder task, it cannot take care of all the tasks. As a result, the translation accuracy of the current neural network machine translation system is not high, and th...

AI Technical Summary

Problems solved by technology

Due to the complexity of the decoder task, it cannot take care of all the tasks, resulting in the current neural network machine

Images