Research on building and utilizing a RAG-based user sentiment analysis avatar AI chatbot

Large Language model(LLM) are becoming a very important tool for conversing with humans in the field of artificial intelligence due to rapid development, and they provide a lot of information to users by learning a vast amount of information in various fields, but companies are trying to build Small Large Language Model(sLLM) tailored to the enterprise by utilizing the information they want. In addition, companies are implementing various ways to communicate with each other. This paper proposes a search augmentation (RAG) algorithm for improving the performance of sLLM construction and an avatar expression change analysis algorithm for user interaction. To evaluate the performance of the algorithm, an enterprise chatbot was built to check the performance.