Sentiment analysis-based social network rumor detection model with bi-directional graph convolutional networks

This paper proposes a model for social network rumor detection that combines sentiment analysis and bi-directional graph convolutional networks (Bi-GCN) to deeply mine the semantic, sentiment, and structural features of information propagation contained in social network texts in order to improve rumor identification’s effectiveness. In this model, a BERT model is used to extract the semantic feature vector from a text, a Bi-GRU+Attention model is used to extract the sentiment feature vector from the text’s comments, and the feature vector is propagated along with the information extracted by the Bi-GCN networks to enrich the rumor detection model’s input features. The experimental results indicate that the precision ratio, recall ratio, and accuracy ratio of the method proposed in this paper are 10%, 9%, and 7% higher than those of the best performing model in the comparison models, respectively, demonstrating the model’s effectiveness.