基于自然语言处理的纠错系统架构设计  

Architecture Design of Error Correction System Based on Natural Language Processing

作　　者：周原 ZHOU Yuan(Schoolof Digital Information Engineering,Minjiang Teachers College,Fuzhou 350018,China)

机构地区：[1]闽江师范高等专科学校数字信息工程学院,福建福州350018

出　　处：《太原师范学院学报（自然科学版）》2022年第3期37-41,46,共6页Journal of Taiyuan Normal University:Natural Science Edition

摘　　要：为提升文本纠错效果,本文研究了一种基于自然语言处理的纠错系统架构.通过建立语言知识库,描述自然语言的上下位关系、同义反义关系等;制定语句合成规则,构建二元句法和三元句法规则集;通过语言分析及文字错误识别功能设计,判断句子中的用词规范性,排查句子中可能存在错误的汉字串;基于自然语言处理构建纠错模型,输入文本,输出概率最大的候选串与原文相结合,得出正确的句子,完成纠错.实验结果证明,应用本文纠错系统架构后,文本错误召回率和正确率均在95.00%以上,对同音词纠错的纠正率为95.76%,长词纠错的纠正率为90.03%,证明本文设计的纠错系统架构具有一定应用价值.In order to improve the effect of text error correction,this paper studies an error correction system architecture based on natural language processing.The language knowledge base is established to describe the superior subordinate relationship,synonymous and antisense relationship of natural language,formulate sentence synthesis rules and construct binary syntax and ternary syntax rule sets,through language analysis and character error recognition function design,judge the standardization of words in sentences,and check the Chinese character strings that may have errors in sentences,the error correction model is constructed based on natural language processing,the text is input and the candidate string with the largest output probability is combined with the original text to obtain the correct sentence and complete the error correction.The experimental results show that after applying the error correction system architecture in this paper,the text error recall rate and correct rate are both above 95.00%,the correction rate for homophones is 95.76%,and the correction rate for long words is 90.03%.The error correction system architecture has certain application value.

关 键 词：自然语言处理 纠错系统 系统架构 知识库构建 句法规则 语义分析 

分 类 号：N945.1[自然科学总论—系统科学]