基于矩阵分解双线性池化的多模态融合虚假新闻检测

现有的大多数虚假新闻检测方法将视觉和文本特征串联拼接，导致模态信息冗余并且忽略了不同模态信息之间的相关性。为了解决上述问题，提出一种基于矩阵分解双线性池化的多模态融合虚假新闻检测算法。首先，该算法将多模态特征提取器捕捉的文本和视觉特征利用矩阵分解双线性池化方法进行有效融合，然后与虚假新闻检测器合作鉴别虚假新闻；此外，在训练阶段加入了事件分类器来预测事件标签并去除事件相关的依赖。在Twitter和微博两个多模态谣言数据集上进行了对比实验，证明了该算法的有效性。实验结果表明提出的模型能够有效地融合多模态数据，缩小模态间的异质性差异，从而提高虚假新闻检测的准确性。

Matrix factorized bilinear pooling for multi-modal fusion fake news detection

Most of the existing methods only connect multi-modal features, without considering the complex connection between modes. To overcome these limitations, this paper proposed an end-to-end algorithm named matrix factorized bilinear pooling for multi-modal fusion fake news detection. Firstly, the algorithm effectively fused the text and visual features captured by multi-modal feature extractor utilizing matrix factorized bilinear pooling method, and then cooperated with fake news detector to classify posts as fake or not. It added the event classifier to remove the event-specific features. This paper compared with this algorithm and the other algorithms on two multimedia datasets, Twitter and Weibo. Experimental results show that the proposed algorithm can effectively fuse multi-modal data, narrow the heterogeneity difference between modes, so as to improve the accuracy of fake news detection.