Construction of an Intelligent Question-Answering System for Xinjiang Geology and Mineral Resources Based on Retrieval-Augmented Generation and Large Language Model

To address the limitations of general-purpose large language models (LLMs) in geology and mineral resources, including insufficient domain knowledge coverage and the tendency to generate “hallucinated” content, this study explores an intelligent question-answering (Q&A) approach tailored to professional applications in Xinjiang geology and mineral resources. Geological survey reports and high-quality journal publications from Xinjiang were systematically collected, and a vision-language model was employed to automatically parse unstructured geological documents, thereby constructing a specialized text corpus. On this basis, an intelligent Q&A system for Xinjiang geology and mineral resources was developed by integrating retrieval-augmented generation (RAG) with LLMs. A text corpus exceeding 23 million characters was established, and the system was locally deployed and evaluated. Experimental results demonstrate that the proposed system significantly outperforms general large language models in professional geological Q&A tasks, markedly improving the professionalism, accuracy, and traceability of the responses. The study effectively compensates for the knowledge deficiencies of general models in specialized domains and provides an innovative technical framework for intelligent services of geological data and mineral resource research in Xinjiang.