基于深度学习的英语语句翻译误差校正方法  

Error Correction Method for English Sentence Translation Based on Deep Learning

作　　者：杨柳青 YANG Liu-qing(School of Humanities,Xi'an Polytechnic University,Xi'an 710048 China)

机构地区：[1]西安工程大学人文学院,陕西西安710048

出　　处：《自动化技术与应用》2022年第12期92-95,共4页Techniques of Automation and Applications

基　　金：2020年校级学位与研究生教育综合改革研究与实践(20yjzg11)。

摘　　要：为提升译文质量,避免翻译歧义,准确呈现源语言内容,以深度学习理念为支撑,面向主流语种构建一种翻译误差校正方法。根据当前词的前几个词,结合多维词向量与各层间的连接权矩阵,建立基于深度卷积神经网络的编码器模型与解码器模型,在解码器模型里添加注意力机制,得到翻译误差校正模型;利用设计的递归神经网络模型生成词向量,依据概率矩阵,采用词对齐模型分割各英语长句为多个短句,按照由目标函数方程得出的最佳调整方案,线性调整译文语序,实现翻译误差校正。经实验验证,所提方法校正效果优势显著,具有较好的有效性。In order to improve the quality of the translation, avoid translation ambiguity, and accurately present the source language content,supported by the concept of deep learning, a translation error correction method is constructed for mainstream languages. According to the first few words of the current word, combined with the multi-dimensional word vector and the connection weight matrix between each layer, the encoder model and decoder model based on deep convolutional neural network are established, and the attention mechanism is added to the decoder model to obtain the translation error correction model;use the designed recursive neural network model to generate word vectors, use the word alignment model to segment each long English sentence into multiple short sentences according to the probability matrix, and adjust the translation linearly according to the best adjustment plan derived from the objective function equation word order, realize translation error correction. The experimental verification shows that the proposed method has significant advantages in the correction effect and has good effectiveness.

关 键 词：深度学习 误差校正 神经网络 连接权矩阵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]