基于边增强一致性与半监督学习的谣言检测研究  

Rumor detection based on edge enhanced consistency and semi-supervised learning

作　　者：张岩珂 但志平 李琳[1,2] 鲁雨洁 ZHANG Yanke;DAN Zhiping;LI Lin;LU Yujie(Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering,Yichang 443002,China;College of Computer and Information Technology,China Three Gorges University,Yichang 443002,China)

机构地区：[1]水电工程智能视觉监测湖北省重点实验室,湖北宜昌443002 [2]三峡大学计算机与信息学院,湖北宜昌443002

出　　处：《现代电子技术》2024年第17期129-135,共7页Modern Electronics Technique

基　　金：国家自然科学基金-新疆联合基金资助项目(U1703261)。

摘　　要：针对现有的谣言检测方法对故意伪造的突发事件检测表现不佳的现象,同时考虑到现实中突发事件的标记数据难以获得,从而导致现有的监督学习方法性能受限,提出基于边增强一致性与半监督学习的谣言检测方法(EECS)。首先通过边增强方法提高数据质量,然后分离出高一致性特征与低一致性特征来深入挖掘内联关系,使用双通道图卷积网络捕获特征,依据半监督学习方法有效利用大量无标记数据增强模型的泛化性,最后采用加权的有监督交叉熵损失和无监督一致性损失优化模型。实验结果表明,在公开的Twitter15、Twitter16和Weibo数据集上,所提出的模型在30%标记样本下准确率达到87.8%、89.5%和95.0%,使用少量标记样本便可达到优异的成绩。The existing rumor detection methods do not perform well in detecting intentionally forged emergencies.At the same time,it is difficult to obtain the labeled data of emergencies in reality,which leads to the limited performance of the existing supervised learning methods.In view of the above,a rumor detection method based on edge enhanced consistency and semi-supervised learning(EECS)is proposed.The edge enhanced method is used to improve data quality,and then,high consistency features and low consistency features are separated from the whole features,so that the internal connection relationships can be explored deeply.A dual-channel graph convolutional network is used to capture features.By the semi-supervised learning method,a large amount of unlabeled data are utilized effectively to enhance the generalization of the model.The model is optimized with weighted supervised cross entropy loss and unsupervised consistency loss.The experimental results show that on the publicly available datasets Twitter15,Twitter16 and Weibo,the proposed model achieves accuracy rate of 87.8%,89.5%and 95.0%under 30%labeled samples,and achieves excellent results with a small number of labeled samples.

关 键 词：谣言检测 半监督 边增强 双通道图卷积 无标记数据 一致性特征 

分 类 号：TN919-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]