摘要
针对电子卷宗文档图像的自动化识别分类对上下文的逻辑结构考虑不充分的问题,提出一种基于神经网络的电子卷宗识别方法。首先,输入电子卷宗图像;然后,利用卷积神经网络识别图像的静态特征;第三,根据电子卷宗的关键要素及上下文关系,将这些静态特征按特定的版面结构转换为时序信号;第四,利用循环神经网络对时序信号进行处理,识别电子卷宗的动态特征;最后,通过对涵盖民事、刑事、行政三类案件类型的电子卷宗的自动识别实验对该方法进行验证。实验结果表明,电子卷宗的自动分类准确率达到95%以上。
In order to solve the problem that the automatic recognition and classification of electronic file image does not consider the logical structure of the context sufficiently,an electronic file recognition method based on neural network is proposed. Firstly,the electronic file image is input,and then the convolutional neural network is used to recognize the static features of the image. Thirdly,according to the key elements and context of the electronic dossier,these static features are converted into timing signals according to the specific layout structure. Fourthly,the cyclic neural network is used to process the timing signals and identify the dynamic characteristics of electronic dossier. Finally,the method is verified by an automatic identification experiment of electronic dossier covering three types of civil,criminal and administrative cases. The experimental results show that the automatic classification accuracy of electronic dossier reaches more than 95%.
作者
李盼
王玉
吴正午
LI Pan;WANG Yu;WU Zheng-wu(China Justice Big Data Institute,Beijing 100043,China;Shanghai Withub General Technology Co.,Ltd.,Shanghai 200232,China)
出处
《中国电子科学研究院学报》
北大核心
2021年第4期363-368,共6页
Journal of China Academy of Electronics and Information Technology
基金
国家重点研发计划资助项目(2018YFC0831600,2018YFC0831603)。
关键词
电子卷宗
神经网络
分类
electronic dossier
neural network
classification
