| 张心源,肖浩宇,李白杨.大语言模型与知识图谱协同增强知识组织理论架构研究[J].数字图书馆论坛,2025,21(5):1~10 |
| 大语言模型与知识图谱协同增强知识组织理论架构研究 |
| Collaborative Enhancement of Knowledge Organization Theoretical Framework by Large Language Model and Knowledge Graph |
| 投稿时间：2024-10-22 |
| DOI：10.3772/j.issn.1673-2286.2025.05.001 |
| 中文关键词: 知识组织；大语言模型；知识图谱；协同增强；语义组织 |
| 英文关键词: Knowledge Organization; Large Language Model; Knowledge Graph; Collaborative Enhancement; Semantic Organization |
| 基金项目:本研究得到教育部人文社科青年基金项目“全球数据主权博弈背景下健全我国数据跨境流动规则体系研究”（编号：21YJC870019）资助。 |
| 作者 | 单位 | | 张心源 | 苏州大学社会学院；苏州大学智能社会与数据治理研究院 | | 肖浩宇 | 郑州大学信息管理学院 | | 李白杨 | 南京大学数据管理创新研究中心 |
|
| 摘要点击次数: 49 |
| 全文下载次数: 134 |
| 中文摘要: |
| 探究了大语言模型（Large Language Model，LLM）与知识图谱协同增强知识组织的有效结合模式，旨在提高知识组织的准确性与效率。依托LLM的自然语言处理与知识生成能力，以及知识图谱对结构化信息的表达和推理能力，提出一种协同增强的知识组织理论架构。首先，系统回顾现有知识图谱应用于知识语义组织的理论与实践发展，梳理LLM在科学知识抽取、实体对齐和图谱融合中的应用，论证知识图谱存储和检索的优化原理，调研LLM与知识图谱协同应用的典型项目。然后，在调研和归纳的基础上，针对知识组织具体操作环节，探究嵌入LLM的合理步骤，重构知识图谱实现知识组织的智能化流程，归纳总结图模协同增强知识组织的有效理论架构。研究发现，图模协同增强可以显著提升知识组织的精确度和可解释性，尤其是在跨领域、多源异构科学知识组织中，能够有效减少LLM的幻觉问题，并提高科学知识检索和问答的准确性与交互性。 |
| 英文摘要: |
| This study explores the collaborative enhancement model of large language model (LLM) and knowledge graph (KG) in knowledge organization, aiming to improve the accuracy and efficiency of knowledge organization. The research combines the natural language processing and knowledge generation capabilities of LLM with the structured expression and reasoning capabilities of KG, proposing a collaborative enhancement method for knowledge organization. First, we systematically review the theoretical and practical development of the existing KG applied to knowledge semantic organization, and sort out the applications of LLM in scientific knowledge extraction, entity alignment, and graph fusion. We demonstrate the optimization principle of KG storage and retrieval, and investigate typical projects of the collaborative application of LLM and KG. Then, based on research and induction, aiming at the specific operation links of knowledge organization, we explore the reasonable steps of embedding LLM, reconstruct the KG to ealize the intelligent process of knowledge organization, and summarize the effective theoretical framework for enhancing knowledge organization through the collaboration of graph and model. The study concludes that the collaboration between graph and model can significantly improve the precision and explainability of knowledge organization, especially in cross-domain, multi-source, and heterogeneous knowledge processing. It effectively reduces the hallucination issues of LLM while enhancing the accuracy and interactivity of scientific knowledge retrieval and knowledge Q&A. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
