[1]李梦洁,董峦.基于 PyTorch 的机器翻译算法的实现[J].计算机技术与发展,2018,28(10):160-163.[doi:10.3969/ j. issn.1673-629X.2018.10.033]
　LI Meng-jie,DONG Luan.Implementation of Machine Translation Algorithm Based on PyTorch[J].,2018,28(10):160-163.[doi:10.3969/ j. issn.1673-629X.2018.10.033]
点击复制

基于 PyTorch 的机器翻译算法的实现()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年10期
页码:
160-163
栏目:
智能、算法、系统工程
出版日期:
2018-10-10

文章信息/Info

Title:
Implementation of Machine Translation Algorithm Based on PyTorch
文章编号:
1673-629X(2018)10-0160-04
作者:
李梦洁; 董峦
新疆农业大学 计算机与信息工程学院,新疆 乌鲁木齐 830000
Author(s):
LI Meng-jie; DONG Luan
School of Computer and Information Engineering,Xinjiang Agricultural University,Urumqi 830000,China
关键词:
机器翻译; 序列对序列; 注意力机制; 词错率; 循环神经网络
Keywords:
machine translation; sequence to sequence; attention mechanism; word error rate; recurrent neural network
分类号:
TP301.6
DOI:
10.3969/ j. issn.1673-629X.2018.10.033
文献标志码:
A
摘要:
当向机器翻译模型输入序列时,随着序列长度的不断增长,会出现长距离约束即输入输出序列的长度被限制在固定范围内的问题,因此所建模型的能力会受到约束。序列到序列模型(sequence to sequence model)可以解决长距离约束问题,但单纯的序列到序列模型无法对翻译中要参考词语前后或其他位置的内容来改善翻译质量的行为进行建模。为了弥补该缺陷,提出了注意力机制(attention mechanism)。针对以上问题,报告了机器翻译及部分模型的研究现状,简述了深度学习框架,分析了基于神经网络的机器翻译及注意力机制原理,并对使用 PyTorch 实现的序列到序列模型及注意力机制进行了研究,通过分析翻译的时间消耗和翻译后的词错率以及评价标准的值来评价模型。最终该模型在英法数据集上取得了一定的效果。
Abstract:
When the sequence is input to the machine translation model,with the increase of sequence length,there will be the problem of long distance constraint,that is,the length of the input and output sequence is limited within a fixed range,so the capacity of the established model will be constrained. The sequence to sequence model can solve the problem of long distance constraint,but the simple sequence to the sequence model cannot model the behavior of improving the quality of the translation in context of translation. In order to make up for the defect,the attention mechanism is proposed. For above problems,we report the research status of machine translation and some models,describe the deep learning framework,analyze the principle of machine translation and attention mechanism based on neural network,and study the sequence to sequence model and attention mechanism implemented with PyTorch. The model is evaluated by analyzing the time consumption of translation,the word error rate after translation and the value of evaluation criteria. Finally,the model has achieved a certain effect on the datasets of English and French.

相似文献/References:

[1]王立霞.面向汉英机器翻译的专利文献小句变换研究[J].计算机技术与发展,2012,(11):77.
　WANG Li-xia.A Chinese-English MT-Oriented Study on Small Sentence Pattern Transformation of Long Patent Sentence[J].,2012,(10):77.
[2]陈家乐,张艳玲.计算机算法类资料的中英文智能翻译[J].计算机技术与发展,2021,31(07):176.[doi:10. 3969 / j. issn. 1673-629X. 2021. 07. 029]
　CHEN Jia-le,ZHANG Yan-ling.English-Chinese Intelligent Translation of Computer Algorithm Corpus[J].,2021,31(10):176.[doi:10. 3969 / j. issn. 1673-629X. 2021. 07. 029]

更新日期/Last Update: 2018-10-10