LLM 기반 진료 정보 조회 챗봇 개발을 위한 프롬프트 설계 및 검증

초록

This study focus on developing an advanced chatbot system that integrates generative AI technologies, such as ChatGPT, with patient medical records to improve telemedicine services, for elderly individuals with limited mobility in adult day care centers. The chatbot employs cutting-edge techniques including Chain of Thought (CoT) prompting and Few-Shot learning, to facilitate accurate and contextually relevant interactions. To valuate its reliability and effectiveness, we create a comprehensive dataset of 478 samples was assessed by medical and technical experts for topic analysis. The evaluation process demonstrated an impressive accuracy of over 89%, showcasing the chatbot's robustness. This achievement highlights the significant potential of the system integrating medical records with prompt engineering.

키워드