Neural networks

Neural network architectures and approaches for fake news detection.

Key papers and articles

• A Deep Learning Approach for Multimodal Deception Detection: Multi-layer perceptron (MLP) variants for multimodal deception detection; demonstrates simple neural architectures with proper feature extraction outperform classical classifiers (SVM, decision trees) by large margins (96.14% accuracy vs. prior 75.20% on courtroom videos).
• Ruchansky, Seo, & Liu (2017) — CSI: A Hybrid Deep Model for Fake News Detection: Combines LSTM networks for temporal pattern capture with fully connected layers for user behavior scoring; demonstrates neural approaches avoid hand-crafted features while achieving state-of-the-art accuracy.
• A Heuristic-driven Uncertainty based Ensemble Framework for Fake News Detection in Tweets and News Articles: Ensemble of multiple pre-trained transformer models with uncertainty estimation via Monte Carlo Dropout; demonstrates complementarity of different neural architectures and importance of uncertainty quantification in fake news detection pipelines.