图书情报工作 ›› 2022, Vol. 66 ›› Issue (24): 84-91.DOI: 10.13266/j.issn.0252-3116.2022.24.008

• 情报研究 • 上一篇    下一篇

学者研究主题演化的时空测度:速度、覆盖度和迂回度

步一1, 黄圣智2, 黄永2, 陆伟2   

  1. 1. 北京大学信息管理系 北京 100871;
    2. 武汉大学信息管理学院 武汉 430072
  • 收稿日期:2022-04-28 修回日期:2022-10-15 出版日期:2022-12-20 发布日期:2022-12-27
  • 通讯作者: 黄永,副教授,通信作者,E-mail:yonghuang1991@whu.edu.cn
  • 作者简介:步一,助理教授,研究员,博士,博士生导师;黄圣智,博士研究生;陆伟,教授,博士生导师。
  • 基金资助:
    本文系教育部人文社会科学研究青年基金项目"复杂网络视角下科学文献的知识融合与知识扩散对比研究"(项目编号:21YJC870001)研究成果之一。

Temporal-Spatial Measurements for Research Topic Evolution of Researchers: Speed, Volume, and Circuitousness

Bu Yi1, Huang Shengzhi2, Huang Yong2, Lu Wei2   

  1. 1. Department of Information Management, Peking University, Beijing 100871;
    2. School of Information Management, Wuhan University, Wuhan 430072
  • Received:2022-04-28 Revised:2022-10-15 Online:2022-12-20 Published:2022-12-27

摘要: [目的/意义]运用深度学习技术,提出结合时间和空间特征的测度(速度、覆盖度和迂回度)方法,用于量化学者研究主题演化,从而为基于内容的学者评价提供量化依据。[方法/过程]提出三维指标框架,其中速度反映作者改变研究主题快慢的平均程度,覆盖度反映作者研究内容所覆盖的主题广度,迂回度反映作者研究路径的曲折性。使用微软学术数据集中计算机科学的作者进行实证研究,并考察学者研究主题演化的三维测度和学者学术影响力和生产力的关系。[结果/结论] 实证研究结果显示,覆盖度与总被引量和总发文量的关系为单调递减,这一特征说明聚焦于特定研究主题较为深入的作者,其发文量和影响力都较大。作者研究主题演化的"速度"和"迂回度"与总被引量、总发文量都存在先增加后减少的倒U型关系。所提出的多维度指标框架不仅可在理论上丰富科学计量学对于学者研究主题转移演化及其机制的理解,而且结合深度学习模型提出了问题的解决思路。

关键词: 研究主题, 深度学习, 学术影响力, 科学评价, 科学计量学

Abstract: [Purpose/Significance] This paper proposes temporal-spatial measurements for research topic evolution of researchers, namely speed, volume, and circuitousness, with deep learning techniques. The measurements are adopted to quantify the research topic evolution of researchers and for quantitatively evaluate researchers based upon their research content. [Method/Process] Among the three indicators, "speed" indicated how fast an author shifts his/her research topics, "volume" implied the width of his/her research topics, and "circuitousness" illustrated how "tortuous" an author’s research topic evolution "path" looked like. This paper adopted authors of computer science recorded in the Microsoft Academic Graph (MAG) dataset and examined the three-dimensional measurement of the evolution of scholars’ research topics and the relationship between scientific impact and productivity of researchers. [Result/Conclusion] Empirical results show that the relationship between volume and total number of citations (total number of publications) is monotonically decreasing, which indicates that authors who focus on specific research topics more deeply have greater publications and influence. Speed and circuitousness of the author’s research topic evolution have an inverted U-shaped relationship (first increase and then decrease) with the total number of citations and the total number of publications. The multi-dimensional indicator framework proposed in this paper not only enriches scientometric understanding of scholars’ research topic shift evolution and its mechanism but also proposes potential solutions to the issues with deep learning-related models.

Key words: research topic, deep learning, scientific impact, science evaluation, Scientometrics

中图分类号: