Article Details

Title · Authors · Abstract · Keywords · DOI · References

MUD-SAPF: A Large-Model-Based Intelligent Parsing and Processing Framework for Multi-Source Unstructured Research Management Data

ZHAO Luxi1, PENG Jie2, TAN Qi1, JI Peng2, LI Lun2

(1.Liangshan Prefecture Science and Technology Information Center, Liangshan 615000, China; 2.Sichuan Institute of Computer Sciences, Chengdu 610041, China)

Section：Artificial Intelligence · Vol. 2026, Issue 17 · Pages：7-12 · 2096-4706（2026）17-0007-06 · CLC number：TP391.4 · Document code：A · DOI：10.19850/j.cnki.2096-4706.2026.17.002

Abstract

Research management involves large volumes of unstructured data, such as credential documents, application guidelines, and project proposals, which traditional information systems struggle to effectively utilize due to heterogeneous layouts, complex semantics, and significant noise. This paper proposes MUD-SAPF, a multi-source unstructured data parsing and structuring framework for research management. By integrating deep layout analysis and multimodal semantic reasoning, the framework establishes an end-to-end data processing pipeline to efficiently structure and clean management data. Experimental results show that MUD-SAPF significantly reduces manual effort and improves recognition accuracy, processing efficiency, and the reliability of downstream analysis in typical application scenarios. The framework demonstrates strong engineering feasibility and supports the intelligent upgrading of research management systems.

Keywords

unstructured data; multimodal large models; research management

References

  1. [1] Chen Y C，Hsieh T C. Big Data for Digital Government: Opportunities, Challenges, and Strategies [J].International Journal of Public Administration in the Digital Age，2014，1（1）：1-14.
  2. [2] Qi M Y，Wang L，LUO Q，et al. IHEP Docs: A Metadata-Based Unstructured Document Management System [C]// International Symposium on Grids & Clouds (ISGC) 2023 in conjunction with HEPiX Spring 2023 Workshop (ISGC&HEPiX2023)，2023：1-14.
  3. [3] Azeroual O. A Text and Data Analytics Approach to Enrich the Quality of Unstructured Research Information [J].Computer and Information Science，2019，12（4）：84-95.
  4. [4] Meng Qingduo, Qin Li, Zong Chaofei Research on Optical Character Recognition Technology and Its Application in Customs Supervision [J]. China Port Science and Technology, 2025, 7 (S1): 63-69
  5. [5] Lin Xin, Du Ying, Luo Yu Structural and functional identification of research project application forms based on multi-stage classification [J]. Digital Library Forum, 2024, 20 (3): 25-33
  6. [6] Rashmi N，Priyadharsini R，Priya Jaculine G. Challenges and Gaps in Scene Text Detection and Recognition: A Detailed Survey [C]//Fifth Congress on Intelligent Systems(CIS 2024)，2025：383-396.
  7. [7] Wang X C. A Review of Text Recognition in Complex Scene Images Based on Deep Learning [C]//Proceedings of the 2025 2nd International Conference on Mechanics, Electronics Engineering and Automation (ICMEEA 2025)，2025：776-783.
  8. [8] Xu Y C，Li M H，Cui L，et al. LayoutLM: Pre-training of Text and Layout for Document Image Understanding [C]//KDD'20: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining，2020：1192-1200.
  9. [9] Kim G，Hong T，Yim M，et al. OCR-Free Document Understanding Transformer [C]//Computer Vision-ECCV 2022，2022：498-517.
  10. [10] Niu J B，Liu Z，Gu Z C，et al. MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing [PP/OL].arXiv.https://doi.org/10.48550/arXiv.2509.22186.
  11. [11] Ateia S，Kruschwitz U，Scholz M，et al. LLM-Based Information Extraction to Support Scientific Literature Research and Publication Workflows [C]//New Trends in Theory and Practice of Digital Libraries (TPDL 2025)，2025：90-99.
  12. [12] Zeng Jun, Wang Ziwei, Yu Yang, etc A review of word embedding methods in the field of natural language processing [J]. Computer Science and Exploration, 2024, 18 (1): 24-43
  13. [13] Bai S，Chen K Q，Liu X J，et al. Qwen2.5-VL Technical Report [PP/OL].arXiv.https://doi.org/10.48550/arXiv.2502.13923.
  14. [14] Wang Z Y，Holmes C. On Subjective Uncertainty Quantification and Calibration in Natural Language Generation [PP/OL].arXiv.https://doi.org/10.48550/arXiv.2406.05213.
  15. [15] Bai Y F，Bao Y P， Charles Y，et al. Kimi K2: Open Agentic Intelligence [PP/OL].arXiv.https://doi.org/10.48550/arXiv.2507.20534.
  16. [16] Chen J L，Xiao S T，Zhang P T，et al. M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation [C]//Findings of the Association for Computational Linguistics: ACL 2024，2024：2318-2335.

Funding

Sichuan Science and Technology Plan Project (2024ZHYS0023); Liangshan Prefecture Science and Technology Plan Project (25YYYJ0220)

About the authors

Zhao Luxi (1990-), female, Han ethnicity, from Huili, Sichuan, assistant researcher, bachelor's degree, research interests include science and technology management; Corresponding author: Li Lun (1992-), female, Hui ethnicity, from Leshan, Sichuan, assistant engineer, master's student, research interests include applied AI research.