Research on developing an offline intelligent system for scientific research data exploitation using large language models and retrieval-augmented generation
DOI:
https://doi.org/10.54939/1859-1043.j.mst.114.2026.175-182Keywords:
Text-to-SQL; Retrieval-augmented generation; Large Language models; Scientific research management.Abstract
This paper presents an offline intelligent framework for exploiting scientific research management databases in secure internal network environments. The proposed system integrates open-source Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) to support natural language interaction with relational databases without requiring SQL expertise. The framework combines three core components: (1) a Text-to-SQL module that converts user queries into executable SQL statements, (2) a RAG mechanism that enriches query generation using database schemas, metadata, and domain knowledge, and (3) locally deployed open-source LLMs optimized for offline inference to ensure data privacy and operational independence. A benchmark dataset comprising approximately 500 natural language–SQL pairs was constructed for evaluation. Experimental results on Qwen3, Meta-Llama, and Gemma demonstrate that the proposed framework achieves reliable SQL generation, efficient inference, and secure offline deployment, thereby improving the accessibility and efficiency of scientific research data exploitation while satisfying stringent security requirements.
References
[1]. S. Adiyono and et al., “AI and Digital Transformation Trends: A Systematic Review with Multi-Criteria Analysis”, Journal of Information Systems and Informatics, Vol. 8, No. 1, pp. 1011–1044, (2026).
[2]. T. Chowdhury, “Cloud-Based Information Retrieval for Big Data: A Survey of Architectures and Scalability Challenge”, The Eastasouth Journal of Information System and Computer Science, Vol. 3, No. 3, pp. 312–322, (2026).
[3]. Z. Hong et al., “Next-generation database interfaces: A survey of llm-based text-to-sql”, IEEE Transactions on Knowledge and Data Engineering, (2025).
[4]. Z. Luo, Z. Yang, Z. Xu, W. Yang and X. Du, “Llm4sr: A survey on large language models for scientific research”, arXiv preprint arXiv:2501.04306, (2025).
[5]. M. Klesel and H. F. Wittmann, “Retrieval-Augmented Generation (RAG)”, Business & Information Systems Engineering, Vol. 67, pp. 551–561, (2025). DOI: 10.1007/s12599-025-00945-3.
[6]. T. Yu, R. Zhang, K. Yang, M. Yasunaga, D. Wang, Z. Li et al., “Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-sql task”, Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pp. 3911–3921, (2018).
[7]. T. H. Nguyen, T. T. Bui and N. H. H. Luong, “A study on deep learning for Vietnamese text classification”, Journal of Military Science and Technology, Vol. 95, pp. 85–94, (2024). DOI: 10.54939/1859-1043.j.mst.95.2024.85-94.
[8]. S. Gupta, R. Ranjan and S. N. Singh, “A comprehensive survey of retrieval-augmented generation (rag): Evolution, current landscape and future directions”, arXiv preprint arXiv:2410.12837, (2024).
[9]. P. P. Mishra, K. P. Yeole, R. Keshavamurthy, M. B. Surana and F. Sarayloo, “A Systematic Framework for Enterprise Knowledge Retrieval: Leveraging LLM-Generated Metadata to Enhance RAG Systems”, arXiv preprint arXiv:2512.05411, (2025).
[10]. L. C. Chen, M. S. Pardeshi, Y. X. Liao and K. C. Pai, “Application of retrieval-augmented generation for interactive industrial knowledge management via a large language model”, Computer Standards & Interfaces, Vol. 94, p. 103995, (2025).
[11]. H. Thakkar and A. Manimaran, “Comprehensive examination of instruction-based language models: A comparative analysis of mistral-7b and llama-2-7b”, 2023 International Conference on Emerging Research in Computational Science (ICERCS), IEEE, pp. 1–6, (2023).
[12]. L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, C. Zhang et al., “Training language models to follow instructions with human feedback”, Advances in Neural Information Processing Systems, Vol. 35, pp. 27730–27744, (2022).
[13]. T. H. Phan, M. P. Nguyen, H. H. Nguyen, T. T. T. Pham and T. L. Le, “Tabular text embedding for Vietnamese text-based person search”, Journal of Military Science and Technology, Vol. 93, pp. 128–136, (2024). DOI: 10.54939/1859-1043.j.mst.93.2024.128-136.
