Method for improving non-autoregressive neural machine translation quality through modeling cooperative relationship

Abstract

The invention discloses a method for improving non-autoregressive neural machine translation quality by modeling a collaborative relationship, and the method comprises the following steps: constructing input of a decoder in a non-autoregressive neural machine translation model by combining source end representation with the length of a target language sequence; obtaining a cooperative relation matrix of the target language sequence by combining the dependency syntax tree, source end representation and decoder input, and finally integrating the cooperative relation matrix of the target language sequence into a decoder in the non-autoregressive neural machine translation model. According to the method, the collaborative relationship between the words in the target sequence is modeled through the dependency syntax tree, and the translation quality is remarkably improved while the dependency relationship is considered.

Description

technical field [0001] The invention relates to the field of neural machine translation models, in particular to a method for improving the quality of non-autoregressive neural machine translation by modeling collaborative relationships. Background technique [0002] With the trend of economic globalization, international exchanges and cooperation have become more frequent. Relying on translators for human translation requires huge manpower and financial resources, which can no longer meet the growing translation needs, and machine translation has emerged as the times require. Machine translation, as the name suggests, refers to the process of using computer technology to convert a source language into a semantically equivalent target language. [0003] Thanks to the improvement of computer computing power and the development of deep learning research, the neural machine translation (Neural Machine Translation, NMT) model based on deep neural network occupies a dominant pos...

AI Technical Summary

Problems solved by technology

[0009] The purpose of the present invention is to provide a method for improving the quality of non-autoregressive neural machine translation models by modeling synergistic relationships, so as to solve the problem of the lack of explicit dependency modeling in existing non-autoregressive neural machine translation models, resulting in reduced translation performance The problem

Images