摘要
变革性研究是科技新范式或新领域产生的前导,识别变革性研究对科研管理和科技前瞻具有重要意义。现有研究较少考虑不同引文关联对目标文献价值判断的影响,为此,本文提出一种融合引文功能和三角引用结构的变革性研究识别方法。根据不同引文功能组合获取目标文献及其前序、后序文献间的三角引用结构,提取文献间的巩固或颠覆关系,据此构建目标文献的自我中心巩固-颠覆引用(ego-centric consolidation-disruption citation,ECCD)网络,以ECCD网络结构特征与文本内容特征为输入,构建图注意力神经网络模型,识别兼具高学术影响力和专家认定颠覆性的变革性研究。在PMCOA(PubMed Central Open Access Subset)数据集上的实证分析发现,变革性研究识别任务的最佳F1值为0.3926,优于其他基线模型。模型参数的可解释性分析显示,本文基于引文功能识别的颠覆性引用关系在高学术影响力研究和变革性研究识别任务中具有重要作用。
Transformative research(TR)serves as a precursor to the emergence of new paradigms or disciplines in science and technology.Identifying TR is important for R&D management and technological forecasting.In response to the gaps in considering how citations of different functions impact the evaluation of focal papers,this study proposed a novel approach to identifying TR by integrating the citation function and triangular citation structure.By obtaining the triangular citation structure among the focal paper,its predecessor,and its successor based on different combinations of citation functions,we extracted the relationships of consolidation or disruption between the papers.An egocentric consolidation-disruption citation network(ECCD)was constructed for each focal paper.The ECCD network structure and text input were employed to build a heterogeneous graph attention neural network model,which was used to identify TR that possessed both high academic impact and peer-reviewed disruptive papers.An empirical analysis of the PubMed Central dataset Open Access subset revealed that the TR identification task achieved an optimal F1-score of 0.3926,which exceeded that of other baseline models.Further parameter analysis showed that disruptive citations interpreted by the citation function played a critical role in identifying high-impact and transformative research.
作者
郑哲浚
马亚雪
梁镇涛
白云
裴雷
Zheng Zhejun;Ma Yaxue;Liang Zhentao;Bai Yun;Pei Lei(Laboratory of Data Intelligence and Interdisciplinary Innovation,Nanjing University,Nanjing 210023;School of Information Management,Nanjing University,Nanjing 210023;School of Information Management,Wuhan University,Wuhan 430072)
出处
《情报学报》
北大核心
2025年第8期950-961,共12页
Journal of the China Society for Scientific and Technical Information
基金
国家自然科学基金青年科学基金项目“基于特征挖掘的科学问题域创新状态建模与突破机理研究”(72204109)
江苏省研究生科研与实践创新计划项目“知识网络视角下基础科学突破识别与机理分析”(KYCX24_0103)。
关键词
变革性研究
引文功能
三角引用
颠覆性
图神经网络
transformative research
citation function
triangular citation
disruption
graph neural network
作者简介
郑哲浚,男,2000年生,博士研究生,主要研究领域为科学计量、数据挖掘;通信作者:马亚雪,女,1993年生,博士,助理教授,博士生导师,主要研究领域为信息组织与知识发现、科技前沿探测,E-mail:mayaxue@nju.edu.cn;梁镇涛,男,1996年生,博士,博士后,主要研究领域为科技情报、知识网络;白云,女,1997年生,博士研究生,主要研究领域为科学社会学;裴雷,男,1981年生,博士,教授,博士生导师,主要研究领域为政策计算与政策扩散、情报安全与数据治理。
