期刊文献+

基于BERT在税务公文系统中实现纠错功能

Implementation of Error Correction Function in Tax Official Document System Based on BERT
下载PDF
导出
摘要 税务公文作为社会政治的产物,具有鲜明的政治性。而撰制公文是一项严肃的工作,必须保持准确、严肃的文体特点。为减轻撰制者和审核者的负担,该实验针对税务系统,利用基于BERT-BiLSTM-CRF的序列标注模型和BERT掩码语言模型的特点,对公文句子中常见的单个字错误进行了检错、纠错实验。准确率、召回率和F1值相比传统的纠错方法有着明显的提升。结果表明,基于BERT-BiLSTM-CRF的序列标注模型和BERT掩码语言模型在税务公文检错纠错应用中具有较大价值。 As a product of social politics,tax official documents have a distinct political nature.Writing official documents is a serious work,and it must maintain accurate and serious style characteristics.In order to reduce the burden of writers and reviewers,this experiment is aimed at the tax system and uses the advantages of the BERT-BiLSTM-CRF-based sequence labeling and BERT mask language model to detect and correct single word errors in official document sentences.Compared with traditional error correction methods,the accuracy rate,recall rate and F1 value are significantly improved.The results show that the BERT-BiLSTM-CRF-based sequence labeling and BERT mask language model have great value in the error detection and correction of tax administrative documents.
作者 袁野 朱荣钊 YUAN Ye;ZHU Rongzhao(School of Computer Science and Information Engineering,Hubei University,Wuhan 430062,China)
出处 《现代信息科技》 2020年第13期19-21,共3页 Modern Information Technology
关键词 税务公文 BERT掩码语言模型 BERT-BiLSTM-CRF 序列标注 tax administrative documents BERT mask language model BERT-BiLSTM-CRF sequence labeling
  • 相关文献

参考文献1

二级参考文献5

  • 1孙才,罗振声.汉语文本校对字词级查错处理的研究[A].第四届计算语言学会议论文集(语言工程)[C].北京:清华大学出版社,1997.319-324.
  • 2张仰森,丁冰青.中文文本自动校对中基于"捆绑+过滤”的字词级查错系统研究与实践[A],1998中文信息处理国际会议论文集[C].北京:清华大学出版社,1998.392-397
  • 3Karen Kukich. Techniques for Automatically CorrectingWords in Text [J] .ACM Computing Surveys, 1992,24(4).
  • 4Andrew R. Golding, Yues Achabes. Combining TrigramBased and Feature-Based Methods for Context-SensitiveSpelling Correction [A]. in: The Computation and Language E-Print Archive[C]. CMP-LG /9605037,1996.
  • 5于勐,姚天顺.一种混合的中文文本校对方法[J].中文信息学报,1998,12(2):31-36. 被引量:22

共引文献3

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部