基于多译本平行语料库的英汉智能翻译系统设计

Design of English-Chinese intelligent translation System based on Multi-translation Parallel Corpus

为了提高传统翻译系统翻译质量,提出一种基于多译本平行语料库的英汉智能翻译系统。为实现该系统,首先采用网络爬虫算法对英汉语料进行收集和预处理,搭建出多译本平行语料库;然后采用基于上下文向量的词对齐模型和基于余弦相似度计算方式的段落对齐模型作为系统模型,并构建出基于attention注意力机制-LSTM的翻译系统,最后与基于跨语言词向量和基于IBM模型1的词对齐模型进行对比实验。实验结果表明,三种模型中基于上下文向量的词对齐模型正确率高达90.24%,而其余两种模型正确率仅为62.71%和51.06%,可以证明基于多译本平行语料库的英汉智能翻译系统可以有效提升翻译的正确率,达到了预期的翻译效果,可以运用于英汉智能翻译的工作中。

In order to improve the translation quality of traditional translation systems,an English-Chinese intelligent translation system based on parallel corpus of multiple translations is proposed.In order to realize the system,firstly,the web crawler algorithm is used to collect and preprocess the English-Chinese corpus and build a multi-translation parallel corpus.Then,the word alignment model based on context vector and the paragraph alignment model based on cosine similarity calculation are used as the system model,and the translation system based on attention mechanism-LSTM is constructed.Finally,the comparison experiment is conducted with the word alignment model based on cross-language word vector and IBM model 1.The experimental results show that the correct rate of the word alignment model based on context vector among the three models is as high as 90.24%,while the correct rates of the other two models are only 62.71%and 51.06%,which proves that the English-Chinese intelligent translation system based on multi-translation parallel corpus can effectively improve the correct rate of translation and achieve the desired translation effect.It can be used in the work of English-Chinese intelligent translation.