摘要
提出一种基于神经网络的翻译预调序方法来预测语言翻译中存在的语序差异,提高外语长句翻译过程中的翻译准确度。通过建立多层神经网络模型,对未标注文本词汇进行向量化处理,实现词汇表示与向量特征的结合,抽取多样本语句和语义的有效信息;在线性排序框架下,采用神经网络进行词语排序评分,获得样本数据语义信息,对语序进行差异预测。通过实验对比表明,采用神经网络预调序模型有效提高了系统性能和翻译准确度。将神经网络翻译模型应用于实际翻译过程中,能降低翻译工作难度,提高翻译效率,具有良好的现实意义。
This paper propose a neural network⁃based translation premodulation method to predict the word order differences in language translation and to improve the translation accuracy of long sentences in foreign languages.Through the establishment of multi⁃layer neural network model,the unlabeled text vocabulary is vectorized,the combination of lexical representation and vector features is realized,and the effective information of multiple statements and semantics is extracted.Under the framework of linear sorting,the neural network is used to sort words,to obtain semantic information of sample data,and to predict the difference of word order.It is shown that the neural network premodulation model can effectively improve the system performance and translation accuracy.Applying the neural network translation model to the actual translation process can reduce the difficulty of translation work and improve the translation efficiency,which has good practical significance.
作者
陈敏
CHEN Min(Xianyang Normal University,Xianyang 712000,China)
出处
《电子设计工程》
2021年第10期24-27,共4页
Electronic Design Engineering
基金
咸阳师范学院专项科研基金项目《译介学视角下中国文化走出去的外译研究——以陕西省为例》的阶段性研究成果(XSYK18032)
咸阳师范学院“青年骨干教师”资助项目(XSYGG201709)。
关键词
机器翻译
神经网络
预排序
差异预测
machine translation
neural network
pre⁃arrangement
variance forecast
