基于融合句法特征的翻译方法研究  

Research on translation methods based on fusion of syntactic features

作　　者：刘晶 LIU Jing(Shaanxi Railway Institute,Weinan 714000,China)

机构地区：[1]陕西铁路工程职业技术学院,陕西渭南714000

出　　处：《电子设计工程》2021年第16期153-157,共5页Electronic Design Engineering

基　　金：陕西铁路工程职业技术学院科研基金项目(KY2019-37);教育部职业院校外语类专业教学指导委员会外语教育教学研究课题(FLEB049)。

摘　　要：机器翻译中的英语长句以其单词较多、句型结构复杂等特征成为影响机器翻译质量的主要因素。文中针对机器翻译中英语长句翻译质量差的问题,提出了基于融合句法特征的翻译方法。该方法结合基于语言模板的翻译方法与基于条件随机场的统计学翻译方法,通过句法维度与统计学维度对长句进行切分及处理,进而提升机器翻译质量。实验结果表明,英语语料集长句在经过文中提出的模型训练后,BLEU分值提升了约2.1%,NIST分值提升了约1.1%,证明了文中所提基于融合句法特征的翻译方法对英语长句翻译质量有所提升。Long English sentences in machine translation are the main factors affecting the quality of machine translation due to their many words and complex English sentence structure.Aiming at the problem of poor translation quality of long English sentences in machine translation,a translation method based on fusion of syntactic features is proposed.This method combines the translation method based on language template and the statistical translation method based on conditional random field to segment and process long English sentences through syntactic and statistical dimensions,thereby improving the quality of machine translation.The experimental results show that the BLEU score of the long English sentence in the corpus is increased by about 2.1%after the model training proposed in the article,and the NIST score is increased by about 1.1%,which proves that the translation method based on the fusion of syntactic features in the article has improved the long English sentence translation quality has improved.

关 键 词：机器翻译 句法特征模型 条件随机场模型 自然语言处理 英语长句翻译 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] H315.9[自动化与计算机技术—计算机科学与技术]