摘要
随着国际交流的增加,有必要对机器翻译模型进行研究,以提高英语翻译的质量。研究开发了一个基于分层先验模型结构的神经机器翻译框架模型,并利用定向动态路由对其进行改进。实验结果表明,FRNN+PRNN模型的翻译性能得到了大幅提升,优化后模型在测试集MT04、MT05、MT06上面的翻译结果分值分别为48.13、45.98、42.85,评分值远远高于RNMT模型和优化前模型。优化后模型在人工和自动评价中的翻译质量分值均最高,具有最高的翻译质量和最少的遗漏、重复翻译;NMT、优化前模型、优化后模型的CDR值分别为0.80、0.76、0.73,说明优化后模型具有很好的翻译忠实度和翻译质量。
With the increase of international communication,it is necessary to study the machine translation model in order to improve the quality of English translation.A neural machine translation framework model based on hierarchical prior model structure is developed and improved by using directed dynamic routing.The experimental results show that the translation performance of the FRNN+PRNN model has been greatly improved.The translation result scores of the optimized model on the test set MT04,MT05 and MT06 are 48.13,45.98 and 42.85,respectively,which are much higher than the RNMT model and the model before optimization.The optimized model has the highest translation quality score in both manual and automatic evaluation,and has good translation fidelity and translation quality.
作者
郭丽娜
GUO Lina(Ankang Vocational and Technical College,Ankang Shaanxi 725000,China)
出处
《自动化与仪器仪表》
2023年第9期192-196,共5页
Automation & Instrumentation
基金
安康职业技术学院课题项目《高职英语教学改革研究》(AZK11009)。
关键词
层次性
模型结构先验
神经机器翻译
英语
hierarchy
model structure prior
neural machine translation
English
context
