人工智能翻译背景下融合多粒度形态特征的神经机器翻译系统研究  

Research on Neural Machine Translation Systems Integrating Multi granularity Morphological Features in the Context of Artificial Intelligence Translation

作　　者：吕千平[1] 刘敏娜[1] LV Qianping;LIU Minna(Xianyang Normal University,Xianyang Shaanxi 712000,China)

机构地区：[1]咸阳师范学院,陕西咸阳712000

出　　处：《自动化与仪器仪表》2024年第5期186-190,共5页Automation & Instrumentation

基　　金：陕西省社科联外语专项《延安精神对外传播英文翻译策略研究》(2022HZ0604)。

摘　　要：在处于经济全球化的今天,神经机器翻译被广泛应用在各种语言的翻译工作中,但其由于语言本身的复杂性和文化背景的差异存在翻译质量低下的问题。因此,研究提出一种多粒度形态特征结合深层编码器信息的融合多粒度形态特征的神经机器翻译系统。研究结果显示,在TOP1与TOP2的评分对比中,融合多粒度形态特征的神经机器翻译模型两项分值均为最高,分别为45.63与49.06。且研究提出的系统平均翻译速度为0.3句/s。综上所述,研究提出的翻译模型与系统能在较短的时间中取得较好的翻译效果,能有效解决目前社会的迫切需求。In today’s era of economic globalization,neural machine translation is widely used in translation work for various languages.However,due to the complexity of the language itself and differences in cultural backgrounds,there is a problem of low translation quality.Therefore,this study proposes a neural machine translation system that combines multi granularity morphological features with deep encoder information to fuse multi granularity morphological features.The research results show that in the comparison of TOP1 and TOP2 scores,the neural machine translation model fused with multi granularity morphological features has the highest scores in both terms,with 45.63 and 49.06 respectively.And the average translation speed of the system proposed in the study is 0.3 sentences/s.In summary,the translation model and system proposed in the study can achieve good translation results in a relatively short period of time and effectively address the urgent needs of the current society.

关 键 词：人工智能 多粒度 形态特征 神经机器翻译系统 深层编码器信息 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]