您的位置：山东大学 -> 科技期刊社 -> 《山东大学学报（理学版）》

《山东大学学报(理学版)》 ›› 2021, Vol. 56 ›› Issue (11): 24-30.doi: 10.6040/j.issn.1671-9352.1.2020.043

• • 上一篇    

基于BERT与法条知识驱动的法条推荐方法

唐光远1,2,郭军军1,2,余正涛1,2,张亚飞1,2,高盛祥1,2   

  1. 1.昆明理工大学信息工程与自动化学院, 云南 昆明 650500;2.昆明理工大学云南省人工智能重点实验室, 云南 昆明 650500
  • 发布日期:2021-11-15
  • 作者简介:唐光远(1992— ),男,硕士研究生,研究方向为自然语言处理.E-mail:1902547647@qq.com
  • 基金资助:
    国家自然科学基金资助项目(61762056,61866020,61761026,61972186);云南省自然科学基金资助项目(2018FB104)

Method of recommendation based on knowledge driven by BERT and law

TANG Guang-yuan1,2 , GUO Jun-jun1,2 , YU Zheng-tao1,2, ZHANG Ya-fei 1,2,GAO Sheng-xiang1,2   

  1. 1. Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunan, China;
    2. Yunnan Key Laboratory of Artificial Intelligence, Kunming University of Science and Technology, Kunming 650500, Yunnan, China
  • Published:2021-11-15

PDF (PC)

325

摘要： 针对传统法条推荐方法知识利用不足的问题,结合预训练BERT模型,提出了一种基于司法领域法律条文知识驱动的法条推荐方法。首先基于BERT预训练模型对法条知识和案件描述分别进行表征,并基于双向LSTM对案件描述文本进行特征提取,然后基于注意力机制提取融合法条知识的案件描述文本特征,最终实现法条智能推荐。该方法在法研杯公共数据集上,法条推荐F1值达到0.88,结果表明,融合法条知识的BERT模型对法条推荐具有显著提升作用,并且可以有效地解决易混淆法条推荐问题。

关键词: 法条推荐, BERT模型, 法条知识融合, 注意力机制

Abstract: Aiming at the problem of insufficient knowledge utilization of traditional law recommendation methods, this article combines the pre-training BERT(bidirectional encoder representation from transformers)model to propose a law recommendation method based on knowledge-driven legal provisions in the judicial field. First based on the BERT pre-training model, the legal knowledge and the case description are characterized separately, and the case description text is extracted based on the two-way LSTM, and then the case description text features combined with the legal knowledge are extracted based on the attention mechanism, and the intelligent recommendation of the legal clause is finally realized. Using this method on the law research cup public data set, the recommended F1 value of the law can reach 0.88. From this effect, it can be seen that the BERT model fused with knowledge of the law can significantly improve the law recommendation and can effectively solve the easily confused method of recommended questions.

Key words: recommendation of law, BERT model, knowledge fusion of law, attention mechanism

中图分类号: 

  • TP391
[1] LAUDERDALE B E, CLARK T S. The supreme courts many median justices[J]. American Political Science Review, 2012, 106(4):847-866.
[2] SEGAL J A. Predicting supreme court cases probabilistic cally: the search and seizure cases, 1962-1981[J]. American Political Science Review, 1984, 78(4): 891-900.
[3] ALETRAS N, TSARAPATSANIS D, PREOTIUC-PIETRO D, et al. Predicting judicial decisions of the European court of human rights: a natural language processing perspective[J]. Peer J Computer Science, 2016. https://peerj.com/articles/cs-93/.
[4] LIU Y H, CHEN Y L, HO W L. Predicting as sociated statutesfor legal problems[J]. Information Processing & Management, 2015, 51(1):194-211.
[5] LONG W, TANG Y, TIAN Y. Investor sentiment identification based on the universum SVM[J]. Neural Computing and Applications, 2018, 30(2): 661-670.
[6] LUO B F, FENG Y S, XU J B, et al. Learning to predict charges for criminal cases with legal basis[C] //Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. Copenhagen: ACL, 2017: 2727- 2736.
[7] JIANG X, YE H, LUO Z C, et al. Interpretable rationale augmented charge prediction system[C] //Proceedings of the 27th International Conference on Computational Linguistics: System Demonstrations. Santa Fe: ACL, 2018: 146-151.
[8] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[EB/OL].(2018-10-11)[2020-05-18]. https://arxiv.org/pdf/1810.04805.
[9] CHEN Q, ZHUO Z, WANG W. BERT for joint intent classification and slot filling[EB/OL].(2019-02-28)[2020-05-18]. https://arxiv.org/pdf/1902.10909.
[10] ADHIKARI A, RAM A, TANG R, et al. DocBERT: BERT for document classification[EB/OL].(2019-04-17)[2020-05-18]. https://arxiv.org/abs/1904.08398?context=cs.
[11] ALBERTI C, LEE K, COLLINS M. A BERT baseline for the natural questions[EB/OL].(2019-01-24)[2020-05-18]. https://arxiv.org/abs/1901.08634.
[12] JI Z C, WEI Q, XU H. Bert-based ranking for biomedical entity normalization[EB/OL].(2019-08-09)[2020-05-18]. https://arxiv.org/abs/1908.03548.
[13] MAO J, LIU W. Factuality classification using the pre-trained language representation model BERT[C] //Proceedings of the Iberian Languages Evaluation Forum(IberLEF 2019). Bilbao: CEUR Workshop Proceedings, 2019: 126-131.
[14] LI W, ZHAO J. TextRank algorithm by exploiting Wikipe dia for short text keywords extraction[C] // Proceedings of the 2016 3rd International Conference on Information Science and Control Engineering(ICISCE). Tokyo: IEEE, 2016: 683-686.
[15] XIAO C J, ZHONG H X, GUO Z P, et al. CAIL 2018: a large-scale legal dataset for judgment prediction[EB/OL].(2018-07-04)[2020-05-18]. https:// arxiv.org/abs /1807.0247.
[16] CAI J, LI J, LI W, et al. Deeplearning model used in text classification[C] // Proceedings of the 2018 15th International Computer Conference on Wavelet Active Media Technology and Information Processing(ICCWAMTIP). Tokyo: IEEE, 2018: 123-126.
[17] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C] // Proceedings of the 31st International Conference on Neural Information Processing. Long Beach: ACM, 2017: 6000-6010.
[1] 鲍亮,陈志豪,陈文章,叶锴,廖祥文. 基于双重多路注意力匹配的观点型阅读理解[J]. 《山东大学学报(理学版)》, 2021, 56(3): 44-53.
[2] 李妮,关焕梅,杨飘,董文永. 基于BERT-IDCNN-CRF的中文命名实体识别方法[J]. 《山东大学学报(理学版)》, 2020, 55(1): 102-109.
[3] 郝长盈,兰艳艳,张海楠,郭嘉丰,徐君,庞亮,程学旗. 基于拓展关键词信息的对话生成模型[J]. 《山东大学学报(理学版)》, 2019, 54(7): 68-76.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2018 《山东大学学报(理学版)》编辑部　
地址: 山东省济南市山大南路27号山东大学中心校区（250100）　电话: +86-0531-88366917　E-mail: xblxb@sdu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发