摘要
智慧司法是智慧城市建设中不可或缺的一部分,智慧司法中法律文书推荐体系的建设完善可以有效解决裁判尺度不统一,类案不同判和量刑不规范等问题.针对现阶段法律文书推荐体系不完善,存在对算力要求高、推荐准确率低和不能满足用户对诉讼请求服务的即时性需求问题,以及为了建立智慧司法中法律纠纷快速响应机制,文中提出了基于深度多模态与核密度估计的法律文书推荐计算模型.首先,构建一个多模态特征融合网络,学习法律文书的融合多模态特征表示;然后,利用核密度估计方法构建类密度距离映射空间;最后,在这个映射空间中通过近邻选择进行法律文书推荐.通过在真实数据集上验证表明,该方法有效减少了推荐误差,提高了推荐准确率.
Smart judicature is an integral part of smart city construction,the construction and improvement of the legal document recommendation system in smart justice makes it possible to effectively solve the problems of inconsistent judgment standards,different judgments in different cases,and irregular sentencing.In view of the imperfect recommendation system of legal documents at this stage,there are drawbacks such as high computing power requirements and low recommendation accuracy.The above problems cannot satisfy the user's demand for real-time litigation request service.In order to establish a quick response mechanism for legal disputes in intelligent judicature,this paper proposes the legal documents recommendation model based on the deep multi-modal and kernel density estimation.Firstly,the multi-view deep networks are trained to learn fusion feature of legal documents.Then,the kernel density estimation method is used to construct the class density distance mapping space.Finally,the legal documents are recommended by selecting the nearest neighbor in the mapping space.Extensive experiments on real-world datasets demonstrate that the proposed method can reduce the recommendation error and improve the recommendation accuracy.
作者
陈志奎
刘振娇
原旭
罗方
赵亮
CHEN Zhi-kui;LIU Zhen-jiao;Yuan Xu;LUO Fang;ZHAO-Liang(School of Software Technology,Dalian University of Technology,Dalian 116620,Liaoning,China;Key Laboratory for Ubiquitous Network and Service Software of Liaoning Province,Dalian University of Technology,Dalian 116620,Liaoning,China)
出处
《西北师范大学学报（自然科学版）》
CAS
北大核心
2021年第1期31-37,共7页
Journal of Northwest Normal University(Natural Science)
基金
国家重点研发计划资助项目(2018YFC0830203)。
关键词
多模态
核密度估计
法律文书推荐
特征融合
近邻选择
multi-modal
kernel density estimation
the legal document recommendation
fusion feature
nearest neighbor
