Fusion of Emotional Information for Rumor Detection Model

Abstract

With the development of technology, the platforms through which people acquire information have shifted from traditional sources such as television and newspapers to today's social media platforms. However, due to the openness of social media platforms, it is challenging to ensure the quality of information. If false information is not addressed promptly, it can adversely affect people's daily lives and lead to social panic. Previous research has largely focused on textual semantic information, which has raised concerns about its limited generalization ability. To address this issue, this study utilized a microblog sentiment analysis dataset to train a sentiment feature extraction model. This trained model was then used for extracting emotional features related to microblog rumors through transfer learning. These emotional features were subsequently integrated with the extracted semantic information features. Experimental results demonstrate that the model achieved an accuracy of 96% on a publicly available rumor detection dataset.