Abstract:

[Purpose] This study aims to build a localized intelligent question-answering system based on Retrieval-Augmented Generation (RAG) technology to address the challenges employees face in accessing information within the complex regulatory framework of the energy industry, and to provide enterprises with an efficient and accurate tool for querying institutional knowledge. [Methods] By integrating the DeepSeek language model with the LangChain framework, this study designs a semantic indexing and generative question-answering subsystem. Through RAG technology, we enhanced the efficiency and accuracy of institutional knowledge retrieval, while supporting enterprise data security protection and dynamic updates. [Results] The system performs well in enterprise regulatory question-answering tasks, with BLEU-1 and ROUGE-1 scores of 0.2651 and 0.3020, respectively, significantly outperforming the baseline without RAG, validating the effectiveness and robustness of the system. [Conclusions] The intelligent question-answering system developed in this study effectively addresses the challenges of accessing institutional knowledge in the energy industry, supports new employee training and enterprise management needs, and provides a reusable technical framework for industry digital transformation.

Key words: large language models, RAG technology, energy industry, intelligent question-answering system