Detecting Health Misinformation on Social Networking Sites Using Large Language Models and Deep Learning-based Natural Language Processing

Detecting Health Misinformation on Social Networking Sites Using Large Language Models and Deep Learning-based Natural Language Processing

연구 분야: Artificial Intelligence

학회: 2024 2nd International Conference on Foundation and Large Language Models (FLLM)

Health misinformation on social networking sites (SNS) is a critical issue, particularly during health crises like the COVID-19 pandemic. The spread of inaccurate health information can lead to severe outcomes, including reduced vaccine uptake and decreased trust in healthcare. This paper introduces a system using large language models (LLMs) and deep learning-based natural language processing (NLP) techniques to detect and mitigate health misinformation on SNS. The model incorporates transformer-based architectures, zero-shot (ZS) and few-shot (FS) learning, and prompt engineering to classify health-related content as true or false. A key feature is the integration of a knowledge graph that enhances verification capabilities. Applied to both English and Arabic datasets, the system’s performance is evaluated using accuracy, F1-score, BLEU, and METEOR metrics, demonstrating its multilingual effectiveness in real-time misinformation detection.