基于双向树多模态融合谣言检测方法的研究  

Rumor Detection Method Based on Bi-directional Tree Multi-modal Fusion

作　　者：马含 李进明 王竹君 关威 MA Han;LI Jin-ming;WANG Zhu-jun;GUAN Wei(School of Computing, Heze University, Heze Shandong 274015,China;School of Information, Shanxi University of Finance and Economics, Taiyuan Shanxi 030006, China)

机构地区：[1]菏泽学院计算机学院,山东菏泽274105 [2]山西财经大学信息学院,山西太原030006

出　　处：《菏泽学院学报》2022年第2期21-25,共5页Journal of Heze University

基　　金：2021年菏泽市社科联委托课题(WT-2021-4)。

摘　　要：对于谣言给社会造成的恐慌,现有谣言检测方法不能有效提高检测精确性.因此,提出一种基于双向树多模态融合谣言检测方法,首先使用预先训练数据向量化,构建端到端的神经网络,以双向树实现特征提取;然后将它输入到全连接层并进行多模态特征拼接,最后采用交叉熵损失函数进行训练.对比传统模型,实验结果表明,该方法能有效提取特征并提高谣言检测的精确性.Rumors cause panic to the society,and the existing rumor detection methods can not effectively improve the detection accuracy.Therefore,a bi-directional rumor detection method based on multi-modal tree is proposed in this paper.Firstly,vectorization of pre-training data is used to construct end-to-end neural network,and feature extraction is realized by bidirectional tree.Then,it is input to the full connection layer for multi-modal feature stitching.Finally,the cross entropy loss function is used for training.Compared with the traditional model,the experimental results show that this method can effectively extract features and improve the accuracy of rumor detection.

关 键 词：深度学习 双向树 多模态融合 谣言检测 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]