Distinguish Confusing Law Articles for Legal Judgment Prediction

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3086–3095 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 3086 Distinguish Confusing Law Articles for Legal Judgment Prediction Nuo Xu1, Pinghui Wang2,1∗, Long Chen1, Li Pan3, Xiaoyan Wang4, Junzhou Zhao1∗ 1MOE NEKEY Lab, Xi’an Jiaotong University, China 2Shenzhen Research School, Xi’an Jiaotong University, China 3School of Electronic, Information and Electrical Engineering, Shanghai Jiao Tong University 4Information Technology Service Center, The Supreme People’s Court, China nxu@sei.xjtu.edu.cn, phwang@mail.xjtu.edu.cn, chenlongche@stu.edu.cn, panli@sjtu.edu.cn, wangxiaoyan@court.gov.cn, junzhouzhao@gmail.com Abstract Legal Judgment Prediction (LJP) is the task of automatically predicting a law case’s judg- ment results given a text describing its facts, which has excellent prospects in judicial assis- tance systems and convenient services for the public. In practice, confusing charges are fre- quent, because law cases applicable to similar law articles are easily misjudged. For address- ing this issue, the existing method relies heav- ily on domain experts, which hinders its appli- cation in different law systems. In this paper, we present an end-to-end model, LADAN, to solve the task of LJP. To distinguish confusing charges, we propose a novel graph neural net- work to automatically learn subtle differences between confusing law articles and design a novel attention mechanism that fully exploit- s the learned differences to extract compelling discriminative features from fact description- s attentively. Experiments conducted on real- world datasets demonstrate the superiority of our LADAN. 1 Introduction Exploiting artificial intelligence techniques to as- sist legal judgment has become popular in recent years. Legal judgment prediction (LJP) aims to predict a case’s judgment results, such as appli- cable law articles, charges, and terms of penalty, based on its fact description, as illustrated in Fig- ure 1. LJP can assist judiciary workers in process- ing cases and offer legal consultancy services to the public. In the literature, LJP is usually formu- lated as a text classification problem, and several rule-based methods (Liu et al., 2004; Lin et al., 2012) and neural-based methods (Hu et al., 2018; Luo et al., 2017; Zhong et al., 2018) have been proposed. The main drawback of existing methods is that they cannot solve the confusing charges issue. ∗Corresponding authors. Judgment results Fact Description Law Articles Charges Terms of Penalty At 18:00 on October 26, 2015, the defendant Zhao XX and Zhang XX had an altercation. Zhao XX beat up Zhang and caused injuries. After identification, the injuries of bilateral nasal bone fractures of Zhang XX were minor injuries of grade ii…… Law Article 234:[The Crime of intentional injury]Whoever intentionally injures another person shall be sentenced to fixed-term imprisonment of not more than three years, criminal detention or public surveillance…… Crime of intentional injury A fixed-term imprisonment of ten months Figure 1: An illustration of the LJP. Generally, a judge needs to conduct professional analysis and reasoning on the fact description of the case, and then choose rea- sonable law articles, charges and the term of penalty to convict the offender. That is, due to the high similarity …

Content truncated to 500 words