Sign in Subscribe
Social media

Enhanced Propaganda Detection in Public Social Media Discussions Using Fine-Tuned Deep Learning Model

Strategic Insights

— 1 min read

Investigations into future internet technologies have highlighted the significance of social media platforms as both vital information sources and conduits for the spread of propaganda and misinformation during the COVID-19 pandemic. A novel approach to detecting propaganda in social media discussions has been proposed, utilizing a fine-tuned deep learning model grounded in the RoBERTa framework. This model demonstrates superior performance in detecting nuanced propaganda cues, surpassing traditional and neural methods in benchmarking tests.

Key Takeaways:

  • The study proposes a novel deep learning framework based on fine-tuning the RoBERTa model for a multi-label, multi-class classification task, which outperforms traditional and neural methods in detecting propaganda cues.
  • The approach offers clear relative advantages, including accuracy, scalability, and contextual adaptability, making it suitable for early adoption by Information Systems researchers and practitioners.
  • The proposed model achieves an overall accuracy of 88%, outperforming state-of-the-art baselines.
  • The study contributes methodological and theoretical insights for combating propaganda in digital discourse, enhancing resilience in online information ecosystems.
  • The research focuses on the diffusion of innovations theory, highlighting the importance of contextual sensitivity in detecting propaganda cues across multiple categories.
  • The study utilizes the RoBERTa model due to its strong contextual representation capabilities and demonstrated superiority in complex NLP tasks.

Statistics:

  • The proposed model achieves an overall accuracy of 88% in detecting propaganda cues.
  • The study benchmarks the proposed model against traditional and neural methods, including TF-IDF, CRFs, and LSTM networks.
  • The RoBERTa model demonstrates superior performance in capturing nuanced propaganda cues across multiple categories.
  • The study utilizes a multi-label, multi-class classification task, which is rigorously benchmarked against traditional and neural methods.
  • The approach is framed within the diffusion of innovations theory, highlighting the importance of contextual sensitivity in detecting propaganda cues.
  • The study contributes to the development of robust, generalizable detection systems in dynamic online environments.

Sources:

  • Enhanced Propaganda Detection in Public Social Media Discussions Using a Fine-Tuned Deep Learning Model: A Diffusion of Innovation Perspective. Future Internet, 2025,17(5):212.
  • https://doi-org.sdpl.idm.oclc.org/10.3390/fi17050212.
  • King Fahad University of Petroleum and Minerals Researchers Add New Findings in the Area of Future Internet (Enhanced Propaganda Detection in Public Social Media Discussions Using a Fine-Tuned Deep Learning Model: A Diffusion of Innovation ...). Medical Letter on the CDC & FDA. June 15, 2025; p 569.

Read more