An Efficient Query System for Coal Mine Safety Information Based on Retrieval-Augmented Language Model

Large Language Models (LLMs) and Retrieval Augmented Generation (RAG) systems hold immense potential for application in industry. In the coal mining, the process of querying and retrieving safety information for dispatchers often consumes considerable time, leading to delays in response to incidents. This paper proposes efficient safety information query system for coal mine safety information. It presents an improvement in the retrieval and generation stages, and establishes a coal mine safety question-answering dataset based on prompt engineering. Various metrics were employed to comprehensively evaluate the RAG framework. The experimental results demonstrate that this work significantly outperform the traditional RAG framework, offering guidelines for RAG implementation in coal mine technical domains.