摘要
针对现有神经网络结构构建的英语机器翻译模型,因长距离依赖导致长距离信息在传递过程中丢失,导致英语机器翻译效果不理想的问题,提出一种注意力嵌入的LSTM英语机器翻译模型。首先,根据标准LSTM网络模型编码阶段采用固定维度向量表示词的特点,在LSTM模型中引入注意力机制,建立了基于LSTM注意力嵌入的英语机器翻译模型;然后,在TensorFlow框架上搭建英语机器翻译系统,并在IWSLT2019数据集上对提出模型进行仿真实验。结果表明,本研究提出的基于LSTM注意力嵌入的英语机器翻译模型,相较于现有神经网络结构构建的英语机器翻译模型,如标准LSTM模型、RNN模型、GRU-Attention翻译模型,可增强源语言上下文信息的表示,提高英语机器翻译模型性能和译文质量。
In order to solve the problem that the existing English machine translation models based on neural network structure lose information in the process of long-distance information transmission due to long-distance dependence, an attention is presented embedded LSTM English machine translation model. Firstly, according to the characteristics of using fixed dimension vector to represent words in the coding phase of standard LSTM network model, attention mechanism is introduced into LSTM model, and English machine translation model based on LSTM attention embedding is established;Then, an English machine translation system is built on tensorflow framework, and the proposed model is simulated on iwslt2019 dataset. The results show that compared with the existing neural network models, such as the standard LSTM model, RNN model and Gru attention translation model, the proposed model can enhance the representation of source language context information and improve the performance and translation quality of English machine translation model.
作者
陈瑞
CHEN Rui(Chongqing Vocational College of Applied Technology,Chongqing 401520,China)
出处
《自动化与仪器仪表》
2021年第10期140-143,共4页
Automation & Instrumentation
基金
重庆市高等教育教学改革项目,名称:同伴互评在高职英语写作教学中的应用研究(No.193532)
重庆市教育委员会人文社会科学研究项目,名称:新时代立德树人视域下高职院校教师职业认同研究(No.20SKGH359)。
关键词
LSTM模型
注意力机制
英语翻译
机器翻译
LSTM model
attention mechanism
English translation
machinetranslation
