Fact Checking Health Claims Using Transformers and Rag Techniques

Abstract - The project aims to build an AI-driven fact checking model for medical and health related claims. The rapid spread of medical misinformation on digital platforms poses a serious threat to public health, leading to misinformed decisions, distrust in scientific research, and potential health crises. This project introduces an AI-powered fact-checking system that verifies the accuracy of health-related claims using advanced NLP techniques. The system leverages BioBERT for medical entity extraction, Retrieval-Augmented Generation (RAG) to fetch relevant evidence from trusted medical sources such as PubMed, WHO, and UMLS, and BERT for claim verification and classification. Claims are categorized based on the retrieved data relevant to the claim and comparing them against it. The BERT model integrated with the project do this and classifies the claims as “Factual,” “False,” or “Insufficient Evidence”, ensuring evidence-backed and real-time verification. To enhance user understanding, the system incorporates a Large Language Model (LLM) that generates contextual explanations, providing insights into the claim’s credibility. The proposed framework automates fact-checking, reducing the reliance on expert verification while improving scalability and accuracy. Unlike traditional methods, which are time-consuming and prone to human bias, this system efficiently processes multiple claims simultaneously, ensuring faster and more reliable fact-checking. Keywords: Fact checking, Health and Medical related claims Verification, Misinformation, Natural Language Processing, Retrieval Augmented Generation, Large Language Models, Evidence-based verification