Research on artificial intelligence assisted decision-making algorithms for lawyers based on legal-computing theory
-
摘要: 针对法学理论和法律实践中缺乏智能决策的问题,综合考虑该领域内的业务数据特征,采用多种数据分析模型进行智能决策算法的研究.法计算学理论以法律关系的数据化智能驱动为核心,在作为法律研究与应用本体的法律关系与计算机科学领域内的数据特征属性之间建立联系,提出了“涵摄分类”概念,并对决策树、朴素贝叶斯等算法进行法律场景下的改进,建立了法律关系坐标系,实现法律关系分析的空间几何转化,最后提出了智能化的辅助决策平台.实验结果表明,该辅助决策与真实律师的办案策略与结果高度吻合,具有辅助律师决策的可行性和有效性.Abstract: At present, there is a lack of intelligent decision-making tools applied to legal theory and practice. Given the characteristics of data in this field, we establish an intelligent decision-making algorithm using a variety of data analysis models. Legal-computing is focused on data-based mechanization of legal reasoning. It establishes a relationship between legal research and applications using the characteristics and data features of computer science. On this basis, the method of "implication classification" is formed, the decision tree and Naive Bayes algorithms are improved for application to the legal arena, and a coordinate system of legal relationships is established to transfer traditional legal relationship analysis into a spatial geometric system. Experimental results show that the algorithm is consistent with a lawyer's handling strategy and results, and has the feasibility of assisting lawyers more broadly in decision-making.
-
Key words:
- legal artificial intelligence /
- legal-computing /
- Naive Bayes /
- C4.5 decision tree
-
表 1 训练数据库的字段与记录结构
Tab. 1 Fields and record structure of training database
字段名 $Req$[0] ${\ldots}$ $Req$[$n$ -1] $Mode$ $Jud$ 首行取值 $V_{Req[0]}^{(1)}$ ${\cdots}$ $V_{Req[n-1]}^{(1)}$ $V_{Mode}^{(1)}$ $V_{Jud}^{(1)}$ 第$k$行取值 $V_{Req[0]}^{(k)}$ ${\cdots}$ $V_{Req[n-1]}^{(k)}$ $V_{Mode}^{(k)}$ $V_{Jud}^{(k)}$ 末行取值 $V_{Req[0]}^{(D)}$ ${\cdots}$ $V_{Req[n-1]}^{(D)}$ $V_{Mode}^{(D)}$ $V_{Jud}^{(D)}$ 表 2 三类典型决策树算法的特性比较
Tab. 2 Comparison of three typical decision tree algorithms
算法 支持模型 树结构 标识特征 特征值类型 树剪枝 ID3 分类 多叉树 信息增益 枚举值 不支持 C4.5 分类 多叉树 信息增益率 枚举值、连续值 支持 CART 分类、回归 二叉树 基尼指数 枚举值、连续值 支持 表 3 实验模型库特征属性与对应模型的记录行数统计
Tab. 3 Statistics on the relationship between characteristic attributes and the number
特征属性 属性取值 $Mode$[0]数 $Mode$[1]数 $Mode$[2]数 $Mode$[3]数 $Mode$[4]数 合计记录数 $Req$[0] 0 0 0 0 0 64 64 1 18 6 3 1 16 44 $Req$[1] 0 8 2 1 0 48 59 1 10 4 2 1 32 49 ${\cdots}$ ${\cdots}$ ${\cdots}$ ${\cdots}$ $Req$[6] 0 9 6 0 1 40 56 1 9 0 3 0 40 52 总计 18 6 3 1 80 108 表 4 与实验案件事实距离相近的法律关系数据点
Tab. 4 Data points of legal relationships close to observations from experimental cases
数据点 $Req$[0] $Req$[1] $Req$[2] $Req$[3] $Req$[4] $Req$[5] $Req$[6] 法律关系模型 欧氏距离 $M$(1) 1 1 1 0 0 0 0 $Mode$[0] 1.00 $M$(2) 1 1 1 1 1 0 0 $Mode$[1] 1.00 $M$(3) 1 0 1 1 0 0 0 $Mode$[1] 1.00 $M$(4) 1 1 0 1 0 0 0 $Mode$[1] 1.00 $M$(5) 0 1 1 1 0 0 0 $Mode$[4] 1.00 表 5 实验样本数据集特征属性与判决结果的记录行数统计
Tab. 5 Statistics of the characteristic attributes and judgement results in the
特征属性 属性取值 $Jud$=0记录数 $Jud$=1记录数 合计记录数 $Req$[0] 0 53 12 65 1 38 30 68 $Req$[1] 0 51 16 67 1 40 26 66 ${\cdots}$ ${\cdots}$ ${\cdots}$ $Req$[6] 0 51 22 73 1 40 20 60 总计 91 42 133 表 6 示例实验数据点及相似数据点的胜诉概率预测结果
Tab. 6 Prediction of winning probability for experimental data points and similar data points
特征要件组合 先验概率 全概率 后验概率 $M$(0) 0.009 45 0.012 95 72.99% $M$(1) 0.008 39 0.013 27 63.24% $M$(2) 0.009 45 0.012 72 74.27% $M$(3) 0.005 68 0.010 14 56.02% $M$(4) 0.002 17 0.007 52 28.90% $M$(5) 0.004 31 0.008 57 43.08% 表 7 决策实验结果汇总记录
Tab. 7 Summary record of decision-making experiments
序号 原特征要件组合 新特征要件组合 胜诉率提升量 决策结果 1 (1, 1, 1, 1, 0, 0, 0) (1, 1, 1, 1, 0, 0, 0) 1.281% 建议变更案由 2 (1, 1, 1, 0, 0, 0, 0) (1, 1, 1, 0, 0, 0, 1) 3.360% 建议变更案由 3 (1, 1, 0, 0, 0, 0, 0) (1, 1, 0, 0, 0, 1, 0) -0.863% 不建议变更案由 4 (1, 1, 0, 1, 0, 0, 0) (1, 1, 0, 1, 1, 0, 0) 1.374% 建议变更案由 5 (1, 1, 1, 1, 1, 0, 0) (1, 1, 1, 1, 1, 0, 1) 2.300% 建议变更案由 6 (1, 1, 0, 0, 0, 1, 1) (1, 1, 0, 0, 0, 1, 0) -2.471% 不建议变更案由 7 (1, 0, 1, 0, 1, 1, 1) (1, 0, 1, 0, 0, 1, 1) -1.647% 不建议变更案由 8 (1, 1, 1, 1, 1, 0, 1) (1, 1, 1, 1, 1, 0, 0) -2.300% 不建议变更案由 -
[1] LEE LOEVINGER. Jurimetrics:The next step forward[J]. Minnesota Law Review, 1949(5):455-493. http://d.old.wanfangdata.com.cn/Periodical/tpyxb201201004 [2] 何勤华.计量法律学[J].法学, 1985(10):38. [3] 屈茂辉, 张杰, 张彪.论计量方法在法学研究中的运用[J].浙江社会科学, 2009(3):21-27. doi: 10.3969/j.issn.1004-2253.2009.03.003 [4] 张妮, 蒲亦非.计算法学导论[M].成都:四川大学出版社, 2015. [5] ATKINSON K, BENCH-CAPON T, MCBURNEY P. Parmendies:Facilitating deliberation in democracies[J]. Artificial Intelligence and Law, 2006(14):261-275. [6] KONING J L, DUBOIS D. Suitable properties for any electronic voting system[J]. Artificial Intelligence and Law, 2006(14):251-260. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=4b447dd7a28e16c70c83e6957e1105d1 [7] 李本.美国司法实践中的人工智能:问题与挑战[J].中国法律评论, 2018(2):54-56. [8] 崔亚东.人工智能与司法现代化[M].上海:上海人民出版社, 2019. [9] HOLMES O. The path of the law[J]. Havard Law Review, 1897(10):457. [10] 宋晖, 刘晓强.数据科学技术与应用[M].北京:电子工业出版社, 2018. [11] 陈为, 朱标, 张宏鑫. BN-Mapping:基于贝叶斯网络的地理空间数据可视分析[J].计算机学报, 2016(7):1281-1293. [12] 韩伟, 沈霄凤, 王云.信息系统的属性重要性度量及知识约简算法比较[J].华东师范大学学报(自然科学版), 2004(3):131-134. doi: 10.3969/j.issn.1000-5641.2004.03.023 [13] 杨青, 王海洋, 卞梦阳, 等.融合贝叶斯推理与随机游走的好友推荐[J].华东师范大学学报(自然科学版), 2018(4):80-89. doi: 10.3969/j.issn.1000-5641.2018.04.008 [14] 匡俊, 唐卫红, 陈雷慧, 等.基于特征工程的视频点击率预测算法[J].华东师范大学学报(自然科学版), 2018(3):77-87. doi: 10.3969/j.issn.1000-5641.2018.03.009 [15] 崔佳旭, 杨博.贝叶斯优化方法和应用综述[J].软件学报, 2018(10):3068-3090. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=rjxb201810010