This is the authors' abstract. We don't add key points for this paper.
The rapid spread of fake news on digital platforms poses a significant challenge to public trust, democracy, and information integrity. Traditional fact-checking methods are often inadequate due to the high volume and speed of misinformation. This paper presents a systematic review of stance detection techniques combined with Natural Language Processing (NLP) for automated fake news detection. We analyze 45 peer reviewed studies (2018–2023) covering machine learning (ML), deep learning (DL), and hybrid approaches, emphasizing transformer-based architectures such as BERT, RoBERTa, and FakeNET. Using benchmark datasets like FNC-1, SemEval-2016, and PHEME, we compare model performance in stance classification and fake news detection, demonstrating that transformer-based models outperform traditional classifiers, achieving up to 97.8% accuracy and a 15% reduction in false positives. However, challenges such as dataset biases, adversarial vulnerabilities, high computational costs, and explainability issues persist. This study highlights the importance of hybrid architectures, adversarial training, and explainable AI (XAI) techniques (e.g., LIME, SHAP) in enhancing fake news detection. Future research should focus on multimodal misinformation detection (text, image, video), cross-lingual stance classification, and resource-efficient NLP models for real-time deployment. By bridging linguistic analysis with computational verification, this review provides a roadmap for scalable, interpretable, and trustworthy fake news detection systems.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Suvidha et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: