Conversational Auto Biography Writing Framework

Recent advancements in Large Language Models (LLMs) have enabled automated natural language processing for complex tasks. This research presents a framework that allows elderly users to create biographies through conversations with a chatbot, preserving their life stories while enhancing their sense of self-worth. The framework consists of two key components: (1) Multi-session RAG Pipeline, which structures interviews into multiple sessions using a chatbot to collect and manage personal narratives efficiently, and (2) Auto Biography Writing System, which generates coherent biographies using a Chain-of-Thought strategy and a Divide and Conquer approach. Experimental results show that the chatbot-based interview system reduces interview fatigue and improves engagement, with a 62% preference rate, while the writing system achieved an 86% win rating, demonstrating its effectiveness. This research has been expanded into the SONJU Project, collaborating with veterans' organizations to apply the framework in real-world settings, confirming its social and historical significance.