Automated Mobile Operator Customer Service Using Large Language Models Combined with RAG System

Large Language Models (LLMs) have made a real breakthrough in the field of artificial intelligence and have been rapidly integrated into our daily lives, including the telecom domain. The use of Retrieval Augmented Generation (RAG) reduces the likelihood of generating incorrect or outdated data in LLMs and improves the understanding of the query context and the generation of more relevant answers. The aim of this research is to improve the quality of automated customer service for mobile operator customers by using RAG system in combination with different language models. The paper makes a comprehensive analysis of open-source LLMs, the main methods of model adaptation and pre-training. Conclusions are drawn on the applicability of the analyses in the study. The architecture of the RAG system and the deployment diagram are designed. The main stages of system training are described, system results and performance evaluation for different LLMs are given. Major conclusions are drawn on the achievement of the research objective and further development.