Analysis of the structure and time-series evolution of knowledge label network from a complex perspective
Xu Wang,Xin Feng,Yuan Guo
DOI: https://doi.org/10.1108/ajim-04-2022-0229
2022-10-06
Aslib Proceedings: New Information Perspectives
Abstract:Purpose The research on social media-based academic communication has made great progress with the development of the mobile Internet era, and while a large number of research results have emerged, clarifying the topology of the knowledge label network (KLN) in this field and showing the development of its knowledge labels and related concepts is one of the issues that must be faced. This study aims to discuss the aforementioned issue. Design/methodology/approach From a bibliometric perspective, 5,217 research papers in this field from CNKI from 2011 to 2021 are selected, and the title and abstract of each paper are subjected to subword processing and topic model analysis, and the extended labels are obtained by taking the merged set with the original keywords, so as to construct a conceptually expanded KLN. At the same time, appropriate time window slicing is performed to observe the temporal evolution of the network topology. Specifically, the basic network topological parameters and the complex modal structure are analyzed empirically to explore the evolution pattern and inner mechanism of the KLN in this domain. In addition, the ARIMA time series prediction model is used to further predict and compare the changing trend of network structure among different disciplines, so as to compare the differences among different disciplines. Findings The results show that the degree sequence distribution of the KLN is power-law distributed during the growth process, and it performs better in the mature stage of network development, and the network shows more stable scale-free characteristics. At the same time, the network has the characteristics of "short path and high clustering" throughout the time series, which is a typical small-world network. The KLN consists of a small number of hub nodes occupying the core position of the network, while a large number of label nodes are distributed at the periphery of the network and formed around these hub nodes, and its knowledge expansion pattern has a certain retrospective nature. More knowledge label nodes expand from the center to the periphery and have a gradual and stable trend. In addition, there are certain differences between different disciplines, and the research direction or topic of library and information science (LIS) is more refined and deeper than that of journalism and media and computer science. The LIS discipline has shown better development momentum in this field. Originality/value KLN is constructed by using extended labels and empirically analyzed by using network frontier conceptual motifs, which reflects the innovation of the study to a certain extent. In future research, the influence of larger-scale network motifs on the structural features and evolutionary mechanisms of KLNs will be further explored.