基于多模态的谣言检测模型研究综述

Analysis of rumor detection algorithm based on multimodality

在互联网时代,由于言论的高自由度,以及各类互联网平台的监管不到位,就容易滋生大量的谣言。相比传统的文本谣言信息,带有图片、视频等多模态信息更加吸引人的注意,同时也增加了谣言检测的难度。文章对多模态谣言检测模型进行归纳分析,将其分为三类:多模态特征融合、多模态特征对比、多模态特征增强。然后介绍了数据集的处理方法,最后对全文进行总结并提出了多模态谣言检测未来的挑战。

In the Internet age,a large number of rumors can easily be generated due to the high degree of freedom of speech and the inadequate supervision of various Internet platforms.Compared with the traditional text rumor information,the modal information with pictures,videos and so on is more attractive and also increases the difficulty of rumor detection.This paper summarizes and analyzes the multimodal rumor detection models,which are divided into three categories:multimodal feature fusion,multimodal feature comparison,and multimodal feature enhancement.Then,the data set processing method is introduced.Finally,the full text is summarized and future challenges of multimodal rumor detection are presented.