Rumor Detection Based on the Temporal Sentiment

Abstract

AbstractThe development of social media has changed the way of information consumption by the public, and it has also shifted the spread of rumors from offline to online. The combined multiscale nature of rumors makes it challenging to develop an effective rumor detection method. This raises the fashion of multi-modal rumor detection. However, these multi-modal methods usually focus on explicit textual and visual features, ignoring the sentiment hidden in textual content, which expresses the individuals’ opinions. Thus, we propose a rumor detection model based on temporal sentiment features in this work. Specifically, we first extract the temporal sentiment feature and text vectors from the text content in normalized reply series, then combine these two vectors as the microblog representation. After that, we apply RvNN to capture the comprehensive representation of the event. Finally, we adopt the multi-layer perceptron neural network to detect rumors. The experiments on two real-world datasets, i.e., Weibo and RumourEval-2019, show that our method performs better than baseline methods. Moreover, the ablation study and the early rumor detection experiments show the effectiveness of our temporal sentiment feature. Our work supplements current rumor detection methods and highlights the important role of temporal sentiment features in rumor spreading.KeywordsRumor detectionRumor spreadingSentiment analyzeTemporal sentimentSocial network