APPLYING HYBRID RAG TECHNOLOGY IN A CHATBOT FOR CONSULTING ON SEXUALLY TRANSMITTED DISEASES (STDS)

Abstract

Sexually transmitted diseases (STDs) remain a sensitive healthtopic, andmany people need trustworthy information that is both accessible andprivacy-preserving. This thesis presents a text-based medical chatbot that provides accurate,confidential, and easy-to-understand STD information. The systemis built usingaRetrieval-Augmented Generation (RAG) architecture with a hybrid retrieverthatcombines lexical search (BM25) and semantic vector search (FAISS). Thetopcandidate passages from both channels are fused using Reciprocal RankFusion(RRF) and injected into the prompt to ground the language model’s answers, withexplicit citations to supporting sources. Experimental results showthat thehybridretrieval strategy improves the relevance of retrieved evidence and enablesmorereliable and transparent responses, while maintaining low end-to-end latencyandlow operational cost suitable for practical deployment. Overall, the proposedprototype demonstrates that a pragmatic hybrid RAG approach can deliver safer,more trustworthy STD guidance and supports the development of user-facingmedical information systems where accuracy and privacy are critical.