Article Details

Abstract

Traditional text information analysis paradigms have fundamental limitations in the deep parsing of complex multimodal documents, which leads to the inability to effectively mine the core information value contained therein. This study aims to utilize the breakthrough capabilities of large models to construct a knowledge fusion framework to achieve structured parsing and cross-modal fusion of multimodal knowledge. This paper designs a closed-loop intelligent framework covering four stages: acquisition, parsing, integration, and fusion. This framework uses large model as the cognitive center to realize intelligent acquisition, deep parsing, high-quality integration, and final cross-modal knowledge fusion of documents. The knowledge fusion framework constructed in this study realizes the systematic transformation from multimodal documents to structured knowledge, and its key efficiency has a significant improvement compared with traditional paradigms. This research framework highly depends on the underlying capabilities of large models and has a large consumption of computing resources. It is necessary to explore the technical paths of model lightweight and inference optimization in the future.

Keywords

multimodal document; knowledge fusion; Large Language Model; information analysis; method system

References

1. [1] Chang Zhijun, Zhang Zhixiong, Qian Li, etc Research and Application Practice of Technology Intelligence Intelligent Data Governance Technology System [J]. Intelligence Theory and Practice, 2024, 47 (1): 22-29
2. [2] Anran, Lu Xiaobin, Zheng Yanning Research on the Method System of Industrial Technology Information Analysis from the Perspective of Digital Intelligence Integration: Analysis Based on Papers in the Field of Library and Information Science [J]. Information Theory and Practice, 2024, 47 (8): 43-52
3. [3] Chen Chunyan Research on Open Source Intelligence Semantic Aggregation Framework Based on Linked Data [J]. Journal of Library and Information Science, 2021, 6 (5): 46-53
4. [4] Yang Qian, Lin He Research on the Automation Framework of Knowledge Fusion Process for Technology Intelligence Basic Services [J]. Information and Management Research, 2024, 9 (4): 65-72
5. [5] Ju Zihan, Wang Yanfei, Bai Rujiang Research on Multimodal Data Fusion in Complex Environments: Taking Open Source Defense Data Analysis as an Example [J]. Intelligence Theory and Practice, 2025, 48 (5): 49-56
6. [6] Wei Wei, Ding Xiangxiang, Guo Mengxing, etc Overview of Text Similarity Calculation Methods [J]. Computer Engineering, 2024, 50 (9): 18-32
7. [7] Lv Xueqiang, Wang Tao, You Xindong, etc Hierarchical Fusion Multi Knowledge Named Entity Recognition Framework - HTLR [J]. Computer Applications, 2025, 45 (1): 40-47
8. [8] Yin Bing, Liu Chenyu, Wei Meng The Path to Advanced Multimodal Visual Ability in Large Models: From Understanding to Interaction [J]. Artificial Intelligence, 2025 (3): 53-63
9. [9] Li Hongliang, Liu Yuliang, Liao Wenhui, etc Optical Character Recognition in the Age of Large Models: Current Status and Prospects [J]. Chinese Journal of Image and Graphics, 2025, 30 (6): 2023-2050
10. [10] Ma Xiaosong, Liu Jie, Li Xiaohui, etc Document layout analysis combining keyword extraction and graph comparison learning [J]. Small and Micro Computer Systems, 2026, 47 (1): 150-156
11. [11] Liu Shu Research and Application of Multi modal Document Knowledge Base Q&A [D]. Shanghai: East China Normal University, 2023
12. [12] Wang Shuai, He Wenchun, Wang Fudi, etc A Question Answering System Integrating Knowledge Graph and Vector Retrieval with Large Language Model [J]. Science, Technology and Engineering, 2024, 24 (32): 13902-13910
13. [13] Dong Jia Internet public document collection system based on vertical search technology [D]. Fuzhou: Fuzhou University, 2016
14. [14] Yang Yutao Research and Implementation of Automatic Grammar Error Correction Model for Chinese Long Text [D]. Chengdu: University of Electronic Science and Technology of China, 2023

Funding

National Key Research and Development Program Cyberspace Security Governance Special Project (2022YFB3103600)

About the authors

Yan Xiaofeng (1978-), male, Han ethnicity, from Xiangtan, Hunan, is a senior engineer with a bachelor's degree. His research interests include the application of big data and artificial intelligence technology, data governance and analysis, and the construction of industrial data platforms; Lang Yujie (1992-), female, Han ethnicity, from Xinyang, Henan Province, engineer, master's degree, research direction: data analysis, system construction, intelligence mining; Yin Lanbing (1989-), female, Han ethnicity, from Xingtai, Hebei Province, engineer, undergraduate, research direction: industrial policy and information research; Huang Chongfu (1990-), male, Han ethnicity, from Qinzhou, Guangxi, is an engineer with a bachelor's degree. His research interests include software architecture design, data mining, and application of AI knowledge management scenarios.