Leveraging Large Language Models with Retrieval-Augmented Generation for an Interactive Slide Presentation System
Ming-Che Chen, Mingshun Wang, Shan Wu, Min-Yu Lin, Chien-Yan Chien, Ren-Guei Hsu, Wan‐Jung Chang, Liang-Bi Chen · 2024
This paper proposes an interactive slide presentation system named Gen-tell-men, based on Large Language Models (LLM) and incorporating the Retrieval-Augmented Generation (RAG) framework. The Gen-tell-men system consists of an edge computing device integrated with a microphone, speakers, and display screen, along with a natural language processing server. The system automatically plays presentation content and responds to visitor inquiries by selecting and displaying relevant slides while generating corresponding voice explanations. Experimental results show that Gen-tell-men is highly efficient in selecting presentation slides related to the inquiries, with an accuracy rate of 0.82, achieving the goal of real-time and appropriate slide presentations and interactive Q&A.