基于深度学习的文本自动纠错系统设计与实现

Design and Implementation of Official Document Automatic Error Correction System

摘要:

为解决机关办公人员对公文格式不熟悉且在公文写作时存在各类格式和内容错误的问题,提出机关公文自动纠错系统,用于辅助公文写作和校对工作；对机关公文纠错需求和常用的纠错方法进行了研究,并进行了机关公文分类标准及公文写作时经常出现格式与内容错误的分析；设计系统由公文模板生成、公文格式校对和公文内容纠错三个功能组成；首先,制作了Word标准格式模板用于实现生成不同类型公文模板；其次,设计公文要素识别与检查算法,并基于VBA技术实现公文格式校对；再次,准备了错别字词查错模型语料库,根据公文规范用语及固定搭配建立了纠错辅助字库,基于Seq2Seq深度学习模型训练字词、语法和标点符号查错模型完成公文内容纠错；最终,通过研究,所设计的系统极大地提升了机关办公人员公文写作效率并减轻校对工作负担。

Abstract:

In order to solve the problems that office staff are unfamiliar with the official document format and there are various errors in the format and content when writing official documents, an automatic error correction system for official documents is put forward to assist the official document writing and proofreading. This paper studied the needs and common error correction methods of official documents, and analyzed the classification standards of official documents and the errors in format and content that often appear in official documents writing. The design system mainly consists of three functions: document template generation, document format proofreading and document content correction. Firstly, the standard format template of Word version was made to generate different types of document templates. Secondly, the algorithm of document element identification and checking was designed, and the document format proofreading was realized based on VBA technology. Thirdly, a corpus of error-checking models for typos was prepared, and an auxiliary word bank for error correction was established according to the standard language and fixed collocation of official documents. Based on the Seq2Seq deep learning model, the error-checking models for words, grammar and punctuation were trained to complete the error correction of official documents. Finally, the system designed greatly improved the efficiency of official document writing of office staff and reduces the burden of proofreading.