基于数据挖掘的自动化翻译系统优化设计研究与应用

Research and Application on Optimization Design of automatic translation System based on Data mining

在全球化和数字化加深的背景下,自动化翻译系统将成为跨语言交流关键。为了克服神经机器翻译精度难以达到需求的问题,研究将在翻译模型本身加入注意力机制以提升精度;同时还对语言本身进行处理基于数据挖掘的处理,以满足各领域的文本信息和句子向量提取结果。结果表明句长处理的聚类神经机器翻译系统性能有较大提升,与标准模型相比BLEU值提升了1.24;单向翻译、双向翻译和多语言翻译情景下性能最优聚类簇头数均为3;尽管与中文蒙语对译相比,中英对译效果更好,但两种语言对译效果均能满足需求,进一步证明了该翻译系统的稳定性。

Under the background of globalization and digitization,automated translation system will become the key to cross-language communication.In order to overcome the problem that the accuracy of neural machine translation is difficult to meet the demand,attention mechanism will be added to the translation model itself to substitute the accuracy.At the same time,the processing of the language itself is based on data mining to satisfy the text information and sentence vector extraction results.The results show that the performance of the cluster neural machine translation system with sentence length processing is improved greatly,and the BLEU value is increased by 1.24 compared with the standard model.The number of optimal cluster in unidirectional,bidirectional and multilingual translation scenarios was 3.Although the translation effect between Chinese and English is better than that between Chinese and Mongolian,the translation effect of both languages can meet the requirements,which further proves the stability of the translation system.