| 李泽宇,刘伟,吴雯娜,过烨琪.基于AI智能体和关键词映射图谱的同义术语挖掘研究[J].数字图书馆论坛,2025,21(1):22~32 |
| 基于AI智能体和关键词映射图谱的同义术语挖掘研究 |
| Synonymous Term Mining Based on AI Agent and Keyword Mapping Graph |
| 投稿时间：2024-10-11 |
| DOI：10.3772/j.issn.1673-2286.2025.01.004 |
| 中文关键词: 同义术语；AI智能体；同义术语挖掘；图论算法；关键词；图谱 |
| 英文关键词: Synonymous Term; AI Agent; Synonymous Term Mining; Graph-Theoretic Algorithm; Keyword; Map |
| 基金项目: |
| 作者 | 单位 | | 李泽宇 | 中国科学技术信息研究所 | | 刘伟 | 中国科学技术信息研究所 | | 吴雯娜 | 中国科学技术信息研究所 | | 过烨琪 | 南京旅游职业学院图书馆 |
|
| 摘要点击次数: 35 |
| 全文下载次数: 87 |
| 中文摘要: |
| 同义术语作为重要的语义资源在信息检索和知识组织等众多领域发挥着重要作用，然而，传统同义术语挖掘方法识别准确度不高且效率低下，难以适应智能化网络时代环境的需要。本文提出使用AI智能体进行同义术语挖掘，基于中文学术文献的中英文关键词映射构建关键词图谱，并提出3种图论算法对存在于同一个关键词图谱的任意两个中文关键词间的同义概率进行量化，从而为AI智能体同义术语挖掘提供辅助参考，实现高效率、精准化同义术语挖掘识别。借助《汉语主题词表》数据对AI智能体进行评估发现，术语关系判断准确率达92.32%，且基于边权连积法对关键词同义概率量化后，量化值前500对关键词数据中同义术语占比近100%，前1 000对关键词数据中同义术语占比超过90%，前1 500对关键词数据中同义术语占比超过80%。实证表明，本文提出的AI智能体和边权连积法相结合的方案可以实现对同义术语的高效率、精准化挖掘发现。 |
| 英文摘要: |
| Synonymous terms, as significant semantic resources, play a vital role in numerous fields such as information retrieval and knowledge organization. However, traditional methods for mining synonymous terms suffer from low accuracy and inefficiency, which are hardly suitable for the demands of the intelligent web era. This paper proposes the use of AI Agent for the mining of synonymous terms. It constructs a keyword map based on the Chinese- English keyword mapping from Chinese academic literature and puts forward three graph-theoretic algorithms to quantify the probability of synonymy between any two Chinese keywords within the same keyword map. This provides auxiliary references for the AI Agent’s mining of synonymous terms, achieving efficient and precise recognition of synonymous terms. Through the evaluation of the AI Agent on the “Chinese Thesaurus” dataset, it is found that the accuracy rate of term relationship judgment reaches 92.32%. Moreover, after quantifying the keyword synonym probability using the edge-weight product method, it is found that synonymous terms account for nearly 100% of the top 500 keyword pairs, over 90% of the top 1 000 keyword pairs, and over 80% of the top 1 500 keyword pairs. Empirical evidence demonstrates that the proposed combination of the AI Agent and the edge-weight product method can facilitate the efficient and precise discovery of synonymous terms. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
