| 优化科学知识图谱方法绘制全领域科学结构图谱 |
| |
| 引用本文： | 陈挺,李国鹏,王小梅.优化科学知识图谱方法绘制全领域科学结构图谱[J].图书情报工作,2022,66(21):107-119. |
| |
| 作者姓名： | 陈挺 李国鹏 王小梅 |
| |
| 作者单位： | 1.中国科学院科技战略咨询研究院 北京 100190;2.中国科学院大学经济与管理学院图书情报与档案管理系 北京 100190;3.中国科学院文献情报中心 北京 100190 |
| |
| 基金项目： | 本文系中国科学院文献情报能力建设专项"重要领域战略情报研究与重点成果"（项目编号：GHJ-QBZX-2021-04）研究成果之一。 |
| |
| 摘 要： | 目的/意义] 针对目前全领域科学知识图谱构建方法中存在的技术难点，结合网络嵌入模型、机器学习聚类、流形学习可视化算法等人工智能领域的方法与模型，提出一套全新发现科学结构的知识图谱构建方案，以完善科学结构发现与可视化布局，并拓展科学知识图谱的分析应用场景。方法/过程] 引入基于深度学习的网络嵌入模型和聚类方法改进原有的网络社团划分聚类方法，利用流形学习降维可视化算法扩大数据处理能力，并设计由下至上分层可视化布局方法，提升可视化图谱的稳定性与细节揭示能力。结果/结论] 以科睿唯安公司的基本科学指标数据库（ESI）研究前沿中高被引论文作为分析数据集，使用新聚类算法得到1 169个研究领域，通过改进的可视化布局算法形成全领域科学结构图谱。与前几期科学结构图谱相比，本文提出的方法支持更大规模的数据分析，对可视化细节揭示与稳定性也有大幅优化，可以更好地展示全领域科学研究宏观结构及内在关系，为全领域科学知识图谱的绘制与构建提供更可靠的方法和技术支持。
|
| 关 键 词： | 科学图谱 科学结构 引文网络 网络嵌入 聚类 可视化 |
| 收稿时间： | 2021-10-11 |
| 修稿时间： | 2022-07-17 |
Optimizing Science Knowledge Mapping Methods to Map the Global Science Structure Graph |
| |
| Chen Ting,Li Guopeng,Wang Xiaomei.Optimizing Science Knowledge Mapping Methods to Map the Global Science Structure Graph[J].Library and Information Service,2022,66(21):107-119. |
| |
| Authors: | Chen Ting Li Guopeng Wang Xiaomei |
| |
| Institution: | 1.Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190;2.Department of Library, Information and Archives Management, School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190;3.National Science Library, Chinese Academy of Sciences, Beijing 100190 |
| |
| Abstract: | Purpose/Significance] To address the technical difficulties in the construction of the current domain-wide science knowledge graph, this study proposes a new knowledge graph construction scheme for discovering scientific structures by combining methods and models in the field of machine learning such as network embedding models, machine learning clustering and manifold learning visualization algorithms, and to improve scientific structure discovery and visualization methods and further expand the analysis application scenarios of science knowledge graph.Method/Process] A deep learning-based network embedding model and clustering method were introduced to improve the original network association partitioning clustering method, a stream learning dimensionality reduction visualization algorithm was used to expand the data processing capability, and a bottom-up hierarchical visualization method was designed to improve the stability and detail revealing capability of the map.Result/Conclusion] Using the Clarivate's Essential Science Indicators (ESI) highly cited papers as the test analysis dataset, 1,169 research areas are obtained by new clustering method, and a global scientific structure map is formed through an improved visual layout algorithm. Compared with the previous science maps, the method proposed in this paper supports larger scale data analysis, and the visualization details reveal and stability are greatly optimized, which can better demonstrate the macro structure and inner relationship of scientific research in the whole domain within a time period, and provide more reliable methods and technical support for the mapping and construction of global science map. |
| |
| Keywords: | science mapping science structure citation network network embedding clustering visualization |
|
| 点击此处可从《图书情报工作》浏览原始摘要信息 |
| 点击此处可从《图书情报工作》下载免费的PDF全文 |
|
