Suppression of false news dissemination on social networks based on multi-modal sentiment analysis

This study proposes a framework for suppressing the spread of fake news on social networks based on multimodal sentiment analysis.This study employs the BERT model to extract contextual semantic vectors from news texts.These are then fused with the output of a bidirectional long short-term memory (BiLSTM) network through feature concatenation, enabling simultaneous capture of local context and global long-range dependencies.Emoticon sentiment features are then extracted through autoencoders and deeply integrated to accurately identify user sentiment inclinations.The study's core innovations are: 1) a multi-tiered fake news detection and suppression architecture; 2) deep fusion of text and emoticon features through multimodal sentiment analysis; 3) dual-strategy dissemination suppression combining 'detection + sentiment immunity'.Experimental results demonstrate that the fake news detection model achieves an accuracy of up to 89.4%.The proposed model can provide effective solutions for building a timely and accurate false news prevention and control system.