Misinformation detection with learning from spatial-temporal propagation features and information content on Twitter
Abstract: This study introduces the STU-User model, an innovative approach to detecting misinformation on social media, combining user behaviors with spatial-temporal analysis. The model incorporates an advanced neural network integrating spatial-temporal units (STU) with enhanced long short-term memory (LSTM) structures. Central to its design is the use of Bert-embeddings to analyze the patterns of users' historical interactions and connections in the network and incorporate them with a similarity measure to enhance accuracy. This analysis is then combined with the spatial-temporal aspects of the message propagation structure. We found that the STU-User model surpasses the performances of existing methods based on tests using public Twitter datasets. Theoretical and practical implications for policy-making and regulating social media in the misinformation era are discussed.
Paper Type: Long
Research Area: NLP Applications
Research Area Keywords: social media, misinformation detection
Contribution Types: NLP engineering experiment, Publicly available software and/or pre-trained models
Languages Studied: English
Submission Number: 714
Loading