RAGQnASystem

RAGQnASystem is an intelligent medical QA system that combines knowledge graph RAG with large language models. It is built upon the DiseaseKG medical dataset, using Neo4j to construct a knowledge graph containing 44,000 entities and 310,000 relationships. Named entity recognition is performed using BERT, while intent recognition and answer generation are handled by a local LLM. By leveraging precise graph retrieval and controlled answer generation, the system effectively mitigates hallucination issues common in large models within medical contexts, significantly improving response accuracy and reliability. Suitable for researchers, doctors, and patients in the healthcare domain.