Intelligent Data-Driven Chatbot for Enhanced Information Retrieval and Interaction

This thesis presents the design and implementation of an intelligent, data-driven chatbot aimed atenhancing information retrieval and diagnostic support within the automotive service domain. Ad-dressing the increasing complexity of modern vehicles and the limitations of traditional diagnosticmethods, the proposed system integrates Retrieval-Augmented Generation (RAG) with LargeLanguage Model (LLM) to create a context-aware, multilingual chatbot. Leveraging structuredand unstructured repair data from Volvo’s workshop records, the chatbot utilizes advanced datapreprocessing, semantic vector search via Azure AI Search, and prompt engineering strategiessuch as Chain-of-Thought reasoning. The system delivers real-time, grounded, and coherentresponses by dynamically retrieving relevant documentation and contextual data. Evaluated us-ing LLM-based metrics for context precision, faithfulness, and answer relevance, the solutiondemonstrates substantial improvements in diagnostic efficiency and user experience. This workcontributes to the growing field of Artificial Intelligence (AI)-powered maintenance tools, highlight-ing the potential of domain-specific conversational agents in complex technical environments.