摘要
[研究目的]近年来网络谣言乱象频出,严重扰乱社会秩序,影响社会稳定。提高网络谣言实时检测方法的准确性,可以帮助公安机关有效治理网络谣言,营造清朗有序的网络环境。[研究方法]提出一种基于领域信息和自定义门控网络的网络谣言实时检测模型CGCRD。模型将多个领域谣言的检测工作视为不同任务,以谣言的内容特征作为主要输入,并补充其领域信息;使用共享专家和独有专家集中学习个性化信息,同时利用多个领域的泛化信息获得更多的信息增益,最终将网络谣言的领域、共有和独有特征加权融合得到嵌入表示,判断是否为谣言。[研究结论]该研究构建了适用于网络安全管理工作的中文多领域谣言数据集RUM-PC23,在该数据集上的实验结果表明,提出的CGCRD模型能够满足公安机关对网络谣言检测工作实时性和准确性的要求,且优于当前的基准模型。
[Research purpose]The spread of online rumors disrupts social order and stability.Improving real-time rumor detection accuracy helps public security agencies manage these rumors and maintain an orderly online environment.[Research method]We propose CGCRD,a real-time detection model based on domain information and customized gate control network.The model treats detection tasks in various domains as separate tasks,using rumor content features and supplemental domain information.Shared and unique experts focus on personalized information,generalized information in different domains are used to generate more information gains.The final embedded representation is obtained by merging domain,shared,and unique features to determine whether the content is a rumor.[Research conclusion]RUM-PC23,a Chinese multi-domain rumor dataset for network security management,is developed.Experimental results show that CGCRD meets the real-time and accuracy requirements of public security agencies and outperforms current baseline models.
作者
安全
徐国天
An Quan;Xu Guotian(School of Public Security Information Technology and Intelligence,Criminal Investigation Police University of China,Shenyang 110854)
出处
《情报杂志》
CSSCI
北大核心
2024年第10期127-133,175,共8页
Journal of Intelligence
基金
辽宁省自然科学基金课题(编号:2022-MS-168)
中国刑事警察学院研究生创新能力提升项目“基于领域特征的网络谣言检测方法研究”(编号:2023YCYB34)研究成果。
