基于改进CMDAM信息检索模型的翻译机器人质量提升研究  

Research on improving the quality of translation robots based on an improved CMDAM information retrieval model

作　　者：王滨[1] 郭黎娜[1] WANG bin;GUO lina(Shaanxi Polytechnic Institute,XianYang Shaanxi 712000,China)

机构地区：[1]陕西工业职业技术学院,陕西咸阳712000

出　　处：《自动化与仪器仪表》2024年第10期278-281,286,共5页Automation & Instrumentation

基　　金：中华职业教育社《职业教育来华留学生高质量培养与赋能“一带一路”国际产能合作效果研究》(ZJS202324)。

摘　　要：为进一步提升英-汉文本翻译场景下的翻译质量,提出一种基于预训练和复制机制的无监督跨语言检索翻译模型,通过实现准确的英汉文本匹配,进一步提升翻译准确率。其中,以基于长短时记忆网络和注意力机制的无监督神经网络机器翻译模型作为基础的检索方法,通过引入预训练和复制机制的方式,进一步提升文本匹配的精度,进而提升翻译质量。实验结果表明,所构建的跨语言检索翻译模型在训练过程中能够迅速达到较低的复杂度和良好的准确率,性能良好且稳定;与其他类型的跨语言文本匹配模型相比,所构建模型在评价指标BLEU上具有更好的表现,准确率更高,BLEU值达到了40.1。综上,构建的基于预训练和复制机制的无监督跨语言检索翻译模型性能优越,可进行准确率良好的英-汉文本匹配,将其应用于英-汉文本翻译场景中时,可进一步提升翻译的准确率。To further improve the translation quality in English Chinese text translation scenarios,a unsupervised cross language retrieval translation model based on pre training and replication mechanisms is proposed.By achieving accurate English Chinese text matching,the translation accuracy is further improved.Among them,the retrieval method based on the unsupervised neural network machine translation model based on long short-term memory network and attention mechanism,further improves the accuracy of text matching and improves translation quality by introducing pre training and replication mechanisms.The experimental results show that the constructed cross language retrieval translation model can quickly achieve low complexity and good accuracy during the training process,with good and stable performance;Compared with other types of cross linguistic text matching models,the constructed model has better performance and higher accuracy in the evaluation metric BLEU,with a BLEU value of 40.1.In summary,the unsupervised cross language retrieval translation model based on pre training and replication mechanism constructed has superior performance and can perform accurate English Chinese text matching.When applied to English Chinese text translation scenarios,it can further improve the accuracy of translation.

关 键 词：信息检索 跨语言文本匹配 英语翻译 机器翻译 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]