黄登娴

摘 要： 传统机器翻译方法采用管道式逐次操作对原始语料实施词性标识以及句法分析，获取英语语言的句法结构，使得翻译任务间存在的错误迭代传递、结构化实例准确性降低，导致英语语言文学翻译准确性降低。因此，对英语语言文学中的机器翻译准确性方法进行校对研究。设计基于知网的词汇语义相似度以及对数线性模型，采用汉英依存树到串的方式保存对应的双语语料，对源语言端实施依存结构化处理，确保汉英双语的对应关系，通过知网运算输入需要翻译句子（依存树结构）同实例库内源语言（依存树结构）中词汇的语义相似度。描述了机器翻译中相似实例检索模块以及译文生成模块的实现过程，通过面向数据的翻译模型进一步校对英语语言的准确翻译。实验结果表明，所提方法可得到准确率高的译文，具有较高的准确性和稳定性。

关键词： 英语翻译软件； 机器翻译； 翻译准确性； 语义相似度； 矫正算法； 迭代传递； 依存树结构

中图分类号： TN912.3?34； TP391.2 文献标识码： A 文章编号： 1004?373X（2018）14?0170?03

Design of translation accuracy correction algorithm for English translation software

HUANG Dengxian

（Civil Aviation Flight University of China， Guanghan 618300， China）

Abstract： In the traditional machine translation method， the syntactic structure of English language is obtained by using the pipeline?type successive operation to perform part of speech identification and syntactic analysis of the original corpus， as a result， wrong iterative transfers exist， and the accuracy of structured examples is reduced in translation tasks， which reduces the accuracy of English language and literature translation. Therefore， a proofreading study is conducted for the machine translation accuracy method in English language and literature. The HowNet?based lexical semantic similarity and logarithm linear model are designed. The Chinese?English dependency?tree?to?string mode is adopted to store the corresponding bilingual corpus， so as to perform dependent structured processing of the source language terminal， and ensure the corresponding relationship between Chinese and English. The lexical semantic similarity degree between sentences （dependency tree structure） to be translated and the source language （dependency tree structure） in the instance base is operated and input by using the HowNet. The implementation processes of the similar instance retrieval module and translation text generation module in machine translation are described. The accurate translation of English language is further proofread by means of the data?oriented translation model. The experimental results show that the proposed method can obtain translation texts with high accuracy rate， and has high accuracy and stability.

Keywords： English translation software； machine translation； translation accuracy； semantic similarity degree； correction algorithm； iterative transfer； dependency tree structure

机器翻译是自然语言操作范围中的关键，具有较高的应用价值。依据实例的机器翻译是一种经验主义的英语语言文学翻译策略，其无需复杂的深层次语法以及语义的分析，提高了英语语言翻译的效率。但是基于实例的机器翻译方法对实例库质量的要求较高[1]。传统机器翻译方法采用管道式逐次操作对原始语料实施词性标识以及句法分析，获取英语语言的句法结构，使得翻译任务间存在的错误迭代传递、结构化实例准确性降低，导致英语语言文学翻译准确性降低。……