电子科技 ›› 2019, Vol. 32 ›› Issue (12): 84-86.doi: 10.16180/j.cnki.issn1007-7820.2019.12.018

• • 上一篇    

基于深度学习的法律文书识别方法研究

孟昕   

  1. 扬州大学 广陵学院,江苏 扬州 225000
  • 收稿日期:2018-12-29 出版日期:2019-12-15 发布日期:2019-12-24
  • 作者简介:孟昕(1990-),女,助教。研究方向:现代高校思政工作及法学信息化。
  • 基金资助:
    江苏高校哲学社会科学研究项目(2019SJB928)

Research on Recognition Method of Legal Documents Based on Deep Learning

MENG Xin   

  1. Guangling College, Yangzhou University, Yangzhou 225000, China
  • Received:2018-12-29 Online:2019-12-15 Published:2019-12-24
  • Supported by:
    Jiangsu University Philosophy and Social Sciences Research Project(2019SJB928)

摘要:

为了提升数字化法律文书知识库的建设效率,文中提出了基于深度学习理论的法律文书识别方法。该方法基于长短期记忆(LSTM)网元结构构建深度神经网络,引入遗忘门进行网元的状态更新,使用Softmax函数作为非线性传播函数,实现自然语言中的实体识别。经测试,该方法可以有效的提取法律文书中的当事人姓名、案由和审判机构等;在文中所采用的测试集上,相较于CRFs算法,该方法在准确率、召回率和F上均可以取得约10%的提升。

关键词: 法律文书, 自然语言处理, 深度学习, 实体识别, LSTM, CRFs

Abstract:

In order to improve the construction efficiency of the knowledge base of digital legal documents, this paper proposed a method of legal document recognition based on the theory of deep learning. This method constructed deep neural network based on long and short term memory (LSTM) network element structure, and realized entity recognition in natural language. After testing, this method could effectively extract the names of people, places and institutions in legal documents. In the test set used in this paper, compared with the CRFs algorithm, this method could achieve about 10% improvement in accuracy, recall and F.

Key words: legal documents, natural language processing, deep learning, entity recognition, LSTM, CRFs

中图分类号: 

  • TN911.7