Purpose: This paper intends to explore methodologies and indicators for the analysis of overlapping structures and evolution properties in a co-citation network, and provide reference for overlapping structure analysi...Purpose: This paper intends to explore methodologies and indicators for the analysis of overlapping structures and evolution properties in a co-citation network, and provide reference for overlapping structure analysis of other scientific networks.Design/methodology/approach: The Q-value variance is defined to achieve overlapping structures of different levels in the scientific networks. At the same time, analyses for time correlation variance and subject correlation variance are used to present the formation of overlapping structures in scientific networks. As a test, a co-citation network of highly cited papers on Molecular Biology & Genetics from Essential Science Indicator(ESI) is taken as an example for an empirical analysis.Findings: Our research showed that the Q-value variance is effective for achieving the desired overlapping structures. Meanwhile, the time correlation variance and subject correlation variance are equally useful for uncovering the evolution progress of scientific research, and the properties of overlapping structures in the research of co-citation network as well.Research limitations: In this paper, the theoretical analysis and verification of time and subject correlation variances are still at its initial stage. Further studies in this regard need to take actual evolution of research areas into consideration.Practical implications: Evolution properties of overlapping structures pave the way for overlapping and evolution analysis of disciplines or areas, this study is of practical value for the planning of scientific and technical innovation.Originality/value: This paper proposes an analytical method of time correlation variance and subject correlation variance based on the evolution properties of overlapping structures, which would provide the foundation for the evolution analysis of disciplines and interdisciplinary research.展开更多
Purpose: The evolution of the socio-cognitive structure of the field of knowledge management(KM) during the period 1986–2015 is described. Design/methodology/approach: Records retrieved from Web of Science were submi...Purpose: The evolution of the socio-cognitive structure of the field of knowledge management(KM) during the period 1986–2015 is described. Design/methodology/approach: Records retrieved from Web of Science were submitted to author co-citation analysis(ACA) following a longitudinal perspective as of the following time slices: 1986–1996, 1997–2006, and 2007–2015. The top 10% of most cited first authors by sub-periods were mapped in bibliometric networks in order to interpret the communities formed and their relationships.Findings: KM is a homogeneous field as indicated by networks results. Nine classical authors are identified since they are highly co-cited in each sub-period, highlighting Ikujiro Nonaka as the most influential authors in the field. The most significant communities in KM are devoted to strategic management, KM foundations, organisational learning and behaviour, and organisational theories. Major trends in the evolution of the intellectual structure of KM evidence a technological influence in 1986–1996, a strategic influence in 1997–2006, and finally a sociological influence in 2007–2015.Research limitations: Describing a field from a single database can offer biases in terms of output coverage. Likewise, the conference proceedings and books were not used and the analysis was only based on first authors. However, the results obtained can be very useful to understand the evolution of KM research.Practical implications: These results might be useful for managers and academicians to understand the evolution of KM field and to(re)define research activities and organisational projects.Originality/value: The novelty of this paper lies in considering ACA as a bibliometric technique to study KM research. In addition, our investigation has a wider time coverage than earlier articles.展开更多
近年来,科学文献呈现增速迅猛、内容复杂、主题细化等特点,给文献分类任务带来了挑战。在此背景下,推动文献自动分类技术的发展,实现科学文献在《中国图书馆分类法》上的正确分类对于信息资源的智能化管理和科学研究的效率化检索具有重...近年来,科学文献呈现增速迅猛、内容复杂、主题细化等特点,给文献分类任务带来了挑战。在此背景下,推动文献自动分类技术的发展,实现科学文献在《中国图书馆分类法》上的正确分类对于信息资源的智能化管理和科学研究的效率化检索具有重要意义。本文提出了多重特征关联和图注意力网络融合的层次分类(hierarchical text clas‐sification networks based on multiple feature correlation and graph attention network,HTCN-MCGAT)模型。该模型由三个模块组成。首先是文献表示与增强模块。为适配文献分类任务,采用表示和增强两阶段流程,重新设计BERT(bi‐directional encoder representation from transformers)预训练模型的微调阶段,使其能够从文献摘要、标题和关键词的内部字符关联以及外部文档关联两个级别实现当前文献的增强表示。其次是标签关联建模模块。使用图注意力网络实现标签语义和层次结构的关系建模。最后是层次交互分类模块。先构建文献和标签的层次融合注意力机制,实现特征空间的文献语义信息与符号空间的层次标签信息的特征关联;再基于多任务学习视角,通过全局和局部信息融合的层次分类网络实现文献分类。本文以中文医学文献作为研究对象,设计系列实验,相较于逐层和平面多分类方法,HTCN-MCGAT模型在F1-score上提高了4.34%~13.21%。此外,还通过样例分析综合验证了本文模型的有效性。本文从特征关联丰富化和层次关系建模两方面对文献分类模型展开优化,在文献分类任务中发挥了较好的应用价值,未来可以推广至更多具有层次结构的分类任务领域。展开更多
基金supported by the National Science Library of Chinese Academy of Sciences
文摘Purpose: This paper intends to explore methodologies and indicators for the analysis of overlapping structures and evolution properties in a co-citation network, and provide reference for overlapping structure analysis of other scientific networks.Design/methodology/approach: The Q-value variance is defined to achieve overlapping structures of different levels in the scientific networks. At the same time, analyses for time correlation variance and subject correlation variance are used to present the formation of overlapping structures in scientific networks. As a test, a co-citation network of highly cited papers on Molecular Biology & Genetics from Essential Science Indicator(ESI) is taken as an example for an empirical analysis.Findings: Our research showed that the Q-value variance is effective for achieving the desired overlapping structures. Meanwhile, the time correlation variance and subject correlation variance are equally useful for uncovering the evolution progress of scientific research, and the properties of overlapping structures in the research of co-citation network as well.Research limitations: In this paper, the theoretical analysis and verification of time and subject correlation variances are still at its initial stage. Further studies in this regard need to take actual evolution of research areas into consideration.Practical implications: Evolution properties of overlapping structures pave the way for overlapping and evolution analysis of disciplines or areas, this study is of practical value for the planning of scientific and technical innovation.Originality/value: This paper proposes an analytical method of time correlation variance and subject correlation variance based on the evolution properties of overlapping structures, which would provide the foundation for the evolution analysis of disciplines and interdisciplinary research.
文摘Purpose: The evolution of the socio-cognitive structure of the field of knowledge management(KM) during the period 1986–2015 is described. Design/methodology/approach: Records retrieved from Web of Science were submitted to author co-citation analysis(ACA) following a longitudinal perspective as of the following time slices: 1986–1996, 1997–2006, and 2007–2015. The top 10% of most cited first authors by sub-periods were mapped in bibliometric networks in order to interpret the communities formed and their relationships.Findings: KM is a homogeneous field as indicated by networks results. Nine classical authors are identified since they are highly co-cited in each sub-period, highlighting Ikujiro Nonaka as the most influential authors in the field. The most significant communities in KM are devoted to strategic management, KM foundations, organisational learning and behaviour, and organisational theories. Major trends in the evolution of the intellectual structure of KM evidence a technological influence in 1986–1996, a strategic influence in 1997–2006, and finally a sociological influence in 2007–2015.Research limitations: Describing a field from a single database can offer biases in terms of output coverage. Likewise, the conference proceedings and books were not used and the analysis was only based on first authors. However, the results obtained can be very useful to understand the evolution of KM research.Practical implications: These results might be useful for managers and academicians to understand the evolution of KM field and to(re)define research activities and organisational projects.Originality/value: The novelty of this paper lies in considering ACA as a bibliometric technique to study KM research. In addition, our investigation has a wider time coverage than earlier articles.
文摘近年来,科学文献呈现增速迅猛、内容复杂、主题细化等特点,给文献分类任务带来了挑战。在此背景下,推动文献自动分类技术的发展,实现科学文献在《中国图书馆分类法》上的正确分类对于信息资源的智能化管理和科学研究的效率化检索具有重要意义。本文提出了多重特征关联和图注意力网络融合的层次分类(hierarchical text clas‐sification networks based on multiple feature correlation and graph attention network,HTCN-MCGAT)模型。该模型由三个模块组成。首先是文献表示与增强模块。为适配文献分类任务,采用表示和增强两阶段流程,重新设计BERT(bi‐directional encoder representation from transformers)预训练模型的微调阶段,使其能够从文献摘要、标题和关键词的内部字符关联以及外部文档关联两个级别实现当前文献的增强表示。其次是标签关联建模模块。使用图注意力网络实现标签语义和层次结构的关系建模。最后是层次交互分类模块。先构建文献和标签的层次融合注意力机制,实现特征空间的文献语义信息与符号空间的层次标签信息的特征关联;再基于多任务学习视角,通过全局和局部信息融合的层次分类网络实现文献分类。本文以中文医学文献作为研究对象,设计系列实验,相较于逐层和平面多分类方法,HTCN-MCGAT模型在F1-score上提高了4.34%~13.21%。此外,还通过样例分析综合验证了本文模型的有效性。本文从特征关联丰富化和层次关系建模两方面对文献分类模型展开优化,在文献分类任务中发挥了较好的应用价值,未来可以推广至更多具有层次结构的分类任务领域。
