TIG-KIGNN: Time Interval Guided Knowledge Inductive Graph Neural Network for Misinformation Detection from Social Media

Since the emergence of social media, misinformation has become prevalent and is propagated through various social media platforms. Time plays a crucial role in verifying the source and authenticity of information, especially when it comes to detecting misinformation. Although current research on misinformation detection often focuses on timestamps for temporal analysis, it tends to overlook the time interval between an event’s occurrence and its reporting, which can provide valuable insights. To address this gap, we propose the Time Interval Guided Knowledge Inductive Graph Neural Network (TIG-KIGNN) for detecting health-related misinformation. Our approach leverages time interval features in social media texts and integrates domain expertise from the knowledge graph into the semantic features of the texts, thereby enhancing the detection process. Moreover, to improve efficiency and minimize resource consumption, we employ inductive graph neural networks to optimize feature representation and update by neighboring nodes during training and when adding new nodes. We validate the effectiveness of our model using a real-world dataset and demonstrate significant improvements over existing methods based on experimental results.