Knowledge mapping of research data in China: a bibliometric study using visual analysis Open Access

Research data (also known as scientific data in China) refers to various types of experimental data, personal observation data, Internet data, statistical data, and simulation data, which are obtained by collection, observation, or analysis, and presented in the form of tables, numbers, images, new media, etc. (Wang, 2018). Research data is both the data source and tool for carrying out scientific research innovation and achieving technical foresight, and forms an important knowledge base that supports the country's decision-making. In recent years, as scientific research has entered an era of intensive data-driven research paradigms, international organizations, governmental departments, and research institutions have all increased their focus on and financial support to the scientific research field. For example, the United Nations' Educational Scientific and Cultural Organization (UNESCO) launched the “Global Alliance for Enhancing Access to and Application of Research Data in Developing Countries,” and the International Council of Scientific Unions (ICSU) established an international organization to promote global research data sharing: “The Committee on Data for Science and Technology and World Data System” (Si and Xing, 2017). The library is a process monitoring and embedded management organization, and an archiving and educational institution for research data, with an irreplaceable position and role in the management, service, and sharing of research data (Sun, 2016).

However, despite its importance, there are very few literature reviews focusing on research data, and those that exist have a limited scope. Brochu and Burns (2019) reviewed some published studies focusing on the relationship between librarians and research data management (RDM). Grant (2017) conducted a literature-based study on the relationship between research data and record-keeping. Ng'eno and Mutula (2018) also studied the core RDM issues in agricultural research institutes. Similarly, Fuhr (2019) investigated a reviewed study on the RDM skills gap among Canadian information workers in the health sciences field. Chawinga and Zinn (2019) investigated and presented a comprehensive account of the factors hampering data sharing at three levels of the global research hierarchy (individual, institutional and international). Hu and Fang (2021) conducted a systematic review of the relevant literature for the evaluation of the research data. Sheng and Yuan (2021) reviews the influencing factors of the openness and sharing of research data from eight aspects: researchers, policy, data, and technology etc. Liu (2020) sorted out the research status of library research data literacy in China, and focused on analyzing the research theme of research data literacy in the library community. Yan et al. (2020) reveals a new paradigm for research data-driven research collaboration and presents future research directions and opportunities. Ruan and Yang (2019) summarized the theories and practices related to information security behavior and research data security management, and commented on the current status and future research direction of research data security behavior. Ma (2019) reviewed the existing research results in the past 10 years from the positioning and function, sharing policy, sharing platform, and sharing strategy, and prospected the future research. The above-cited literature review studies indicate that yet, there is no systematically organized research study that covers all the important aspects of research data.

Research related to research data is an interdisciplinary field, and researchers encounter different issues, according to their respective knowledge backgrounds. Before the advent of bibliometric tools, researchers relied on peer-reviewed articles or documents to quickly obtain a panoramic view of a subject or field of research (Chen and Chen, 2017). Some obvious limitations of this approach are that the research results are ultimately influenced by the vision and subjective judgment of the peers' own knowledge, it cannot completely reveal the critical studies in the field or the emerging research hotspots, and it is controversial. Bibliometric tools offer researchers another possibility however, where the literature review changes from a qualitative research method to a mixed research method, thereby leading to more objective and reliable research results.

This study combined bibliometric research methods with knowledge maps, and then systematically reviewed the published studies on research data. This study provides a panoramic view of research data, and through library informatics, offers a quick understanding of the researchers' collaborative characteristics, institutional collaboration characteristics, hot research topics, evolutionary trends, and research frontiers. The specific research questions are as follows: (1) What are the collaboration characteristics between authors and institutions on the topic of research data in the Chinese library and information field. (2) What are the major research subjects relating to “research data” in the Chinese library and information field? (3) What evolutionary paths have the research hotspots and cutting-edge information followed?

Many scholars have used various evaluation methods to try and carry out quantitative analyses and so gain influence for the authors and research institutions. They have also proposed many valuable measurement indicators and evaluation methods, which can be mainly placed in two categories: “Based on Bibliometric Analysis” and “Based on Social Network Analysis” (Chao et al., 2016). The evaluation method using a bibliometric analysis is based on indicators such as the number of publications, CiteSpace and VOSviewer are both important tools for information visualization and bibliometric knowledge map research in recent years, but they are also different in the theoretical algorithm of visual map generation. CiteSpace focuses on expressing the strength of the relationship by graphics and connections, while VOSviewer mainly calculates the relationship by distance. CiteSpace software has certain advantages in revealing the dynamic development law of the discipline and discovering the research frontier. VOSviewer software performs better when the relationship between subject themes is clearly presented, or when the amount of data is very large.

CiteSpace software employs a cosine algorithm to calculate the cooperation intensity of the researchers or institutions. The connection strength between nodes represents the cooperation strength between the researchers or institutions. This is calculated by the cosine distance of the angle between the nodes. Formula (1) is as follows:

where $cij$ represents the number of papers published by the co-authors (author i and author j), $Si$ and $Sj$ represent the number of papers published by author i and author j, respectively, and the value of the cooperation strength is between 0 and 1.

VOSviewer uses the correlation strength algorithm, as shown in formula (2):

In formula (2), $Cij$ represents the number of papers published by the co-authors (author i and author j), $Wi$ and $Wj$ represent the number of papers published by author i and author j respectively, and $Sij$ represents the similarity between author i and author j. It should be noted that the accuracy of VOSviewer's association strength algorithm can be guaranteed only if author i and author j are independent of each other. Therefore, the association strength algorithm measures the similarity from the perspective of probability.

We use CiteSpace and VOSviewer to generate different maps and compare them. It is found that the maps generated by CiteSpace have richer colors and more beautiful appearance. In addition, we can view the articles involved in a node, the scale and content of clustering, and the average year of clustering from the map. Therefore, we decided to use CiteSpace to analyze the data of this study. By using CiteSpace, this paper is able to draw visual knowledge maps, and obtain the cooperative relationships present between the author and research institutions, as well as identify the research trends in the field of research data.

China National Knowledge Infrastructure (CNKI) is the largest continuously and dynamically updated full-text database of Chinese academic journals globally, and it is the most authoritative document retrieval tool and network publishing platform for Chinese academic journals. It contains all the academic journals in China and covers the contents of all disciplines. Many databases such as CSSCI, SCI, master's degree papers, doctoral theses are in CNKI. The subject of the paper is Chinese research data, which belongs to the library and information discipline. All the published documents on this subject are included by CNKI. With the data source of CNKI, this paper adopted the advanced professional search, with search formula of Subject = ‘data sharing’ + ‘data management’ + ‘data curation’ + ‘scientific data’ + ‘research data’ + AND Subject = ‘library’, to initially research and obtain 1,260 literature records (the search time was December 13, 2019). The author imported the searched literature into CiteSpace software to automatically check the weight and then manually eliminated the requirements for manuscript collection, forum notification, news reports, popular science essays, and other non-research documents. Finally, 1,238 documents were determined.

Scientific research collaboration was defined by scientific metrologists Katz and Martin as follows: research scholars work together for common purposes of jointly producing new scientific knowledge. The attributive information of scientific research collaboration was mainly derived from the research of published authors. Therefore, the author of this paper introduced the data into the software with the authors as the nodes, selected the time span from 1988 to 2018, of which the time slice was four years, and the threshold was top 50 publications at each stage, to carry out the visualization analysis, and finally obtained Figure 1. Moreover, the authors of the top ten publication amounts were listed in Table 1. In Figure 1, the node represents the author, the connecting line represents the partnership, and the thickness of the connecting line represents the strength of the relationship.

It can be concluded from Figure 1 that Gu (2018) corresponds to the largest node and has published 13 articles in the field, mainly focusing on the policies of open access to scientific research data and the rights and interests of data management services. Shen (2015) ranks second in the corresponding node and has published 11 articles, mainly focusing on supervision method of research data, librarians' data literacy connotation, and training system. The statistics in Table 1 show that the top ten authors are in relatively important positions in this field. Their studies can help researchers quickly understand the current status and development of research data. From the perspective of collaborative relationship, there are nine scholars in the largest collaborative area, mainly Hu (2015). The main research content of this team is the process supervision of biomedical data in the era of big data. The collaboration team, ranking second, consists of 7 scholars, including Meng et al. (2016), who have mainly studied the data management system and data literacy of the library. The collaboration team headed by Hu (2015) ranks third, which takes chemistry as an example to discuss relevant policies and services for the publication of scientific research data in subject areas and studies data literacy education in the libraries of foreign universities. Moreover, this team includes three scholars among the top ten publications, indicating that the team is one of the key high-performance teams in the research data field from the library perspective.

For CiteSpace's structural control of the network graphs, the k network can be set to filter out the smaller network structures (Note: k refers to the top k largest network structures); when k = 1, the obtained network is the largest subnet of the graphic structure. In order to further clarify the collaborative intensity between the research scholars, the author screened the information in Figure 1 and set to display the top five subnet structures and the connecting strength, as shown in Figure 2. Among five largest subnet graph structures, the collaborative strength is 0.84 between Wu and Hu (2016), 1.0 between Jie and Sheng (2016), 0.5 between Meng and Qian (2013), 1.0 among Li et al. (2014), and 0.61 between Shen and Hao (2016), these are the collaborative strengths of the largest collaboration teams in the field. In summary, the collaborative relationship is relatively strong in values in this field, but the number of publications is low, which should be strengthened further.

In order to analyze the characteristics of the research institutions of the library's research data, the node was set as the operation of institutions to obtain the institutional collaboration graph as shown in Figure 3, and the institutions of top 10 publication amounts were selected to form Table 2. It can be seen from Figure 3 and Table 2 that Wuhan University ranks first and has published the most articles (totally 68) in this field, including School of Information Management, Library, and Center for the Studies of Information Resources of Wuhan University; Shanghai University Library has published 38 articles, ranking second; the Documentation and Information Center of Chinese Academy of Sciences has published 28 articles, ranking third, which is then followed by University of Chinese Academy of Sciences (25 articles), National Science Library of Chinese Academy of Sciences (13 articles), Southeast University Library (12 articles), Medical Library of Chinese PLA (12 articles), National Library (12 articles) and Dept. of Information Management of Nanjing University (11 articles) among top ten publication amounts. It can be seen that the initial publication time was 2004, the earliest, for the School of Information Management of Wuhan University, 2007 for Center for the Studies of Information Resources of Wuhan University, 2009 for National Science Library of Chinese Academy of Sciences, 2010 for National Library, 2012 for Southeast University, Shanghai University Library, Medical Library of Chinese PLA and Dept. of Information Management of Nanjing University, and 2014 for Documentation and Information Center of Chinese Academy of Sciences and University of Chinese Academy of Sciences. So, Wuhan University Library, Shanghai University Library and Documentation and Information Center of Chinese Academy of Sciences are high-performance institutions in this field. Meanwhile, the School of Information Management of Wuhan University took the first step early, while the Chinese Academy of Sciences started late but has made very rapid progress in research.

The keywords are concise summaries of the topics and contents of the literature research. It is helpful to know the basic research contents of the literature via correct analysis of the keywords and know the essential hot topics of the subjects, institutions, and research knowledge in a certain period by measuring the number of the keywords (Zhao and Jiang, 2014). In this paper, the author set the node as the keyword, selected the period of 1988–2019 with a slice of 4 years and the top 50 frequent keywords in each stage for visualization, adopted the minimum spanning tree MST to prune the generated graph, and finally clustered the results and extracted them with k (keyword) as the label to obtain Figure 4. In order to visually make out the corresponding frequency and centrality of the keywords, Table 3 was prepared by selecting the top 30 frequent keywords. Clustering was achieved by layering the intimacy and similarity between the research data from high to low. The structure and clearness of CiteSpace clustering were mainly determined by two indicators: modularity (Q-value) and average silhouette (S-value for short). The larger the Q-value was, the better the clustering of the network became; moreover, the Q-value interval was [0, 1]; Q > 0.3 indicated that the clustering network structure was significant. The S-value could be used to measure the homogeneity of the clustering graph; when it was approaching 1, the homogeneity was higher; when it was above 0.5, it was considered that the clustering result was reasonable. As S = 0.6022 in Figure 4, it was judged that the clustering structure obtained in this study was clear, and the result was very reliable.

In the graph, each node corresponds to a keyword. The connecting line indicates the co-occurrence relationship between the corresponding keywords, e.g. the co-occurrence relationships between the digital library and cloud computing and between university library and data literacy. The purple edge refers to the point with high intermediate centrality (centrality ≥ 0.1), which is generally considered as a pivot node, such as big data, digital library, data literacy, data management, etc. The flow of knowledge can be judged by the color of each annual ring, e.g. time (color) (the transfer from cool color to warm color means the temporal variation from far to near) (Li and Chen, 2015). It can be visually seen from the graph that the hot research topics in this field are mainly #0 data literacy, #1 scientific data, #2 information resource sharing, #3 joint cataloging, #4 resource sharing, #5 data integration, #6 research support, #7 digital library, etc., of which top five topics will be analyzed in detail in the following paragraphs, the research data, as the subject of this study, will not be discussed further, and #2 information resource sharing and #4 resource sharing are summarized as research data sharing.

Research data sharing includes resource collaboration, joint construction and sharing, and mutual coordination to meet the needs of users' research activities up to the hilt between university libraries and public libraries or other institutions. #3 scientific data sharing mainly includes 15 keywords, such as digital resources (16, 0.10), library alliance (10, 0.06), document information resources (5, 0.03), information resource sharing system (2, 0.00), data warehouse (4, 0.07), etc. of which the single contour value is 0.753; #4 resource sharing mainly includes 14 keywords, such as cloud computing (42, 0.10), resource sharing (46, 0.18), joint construction and sharing (24, 0.05), literature resources (4, 0.02), etc., of which the single contour value is 0.81. In the era of big data, driven by the data-intensive paradigm, research data sharing has been highly valued in the field of library and information. Among the others, Wang et al. (2008) adopted bibliometrics to analyze the articles of research data sharing from time, journal, and topic in China. Wei and Zhu (2007) and Huang et al. (2009) analyzed several measures for the participation of academic libraries in research data sharing. Zhang (2017) discussed the data sharing mode of university libraries from three levels of data research and development, data collection, and data usage. Si and Wang (2018) investigated six research data platforms under the National Basic Science and Technology Condition Platform Project and described the current data organization situation, existing problems, and improvement suggestions for the platforms. In general, the studies on research data sharing in the domestic library and information field mainly focus on dynamic analysis of librarian's research data sharing, dialectical relationship between libraries and research data sharing, new technology of libraries participating in research data sharing platform, and research data sharing mode and practice of libraries.

Data literacy is an extension of information literacy. It mainly includes three aspects of data consciousness, data capabilities (collection and processing, representation and description, discovery and retrieval, selection and evaluation, analysis and utilization, integration and reuse, preservation, and management). Throughout the data lifecycle, data ethics is one of the essential attainments of people in the E-science environment (Huang and Li, 2016). The #0 data literacy mainly includes 21 keywords, including college library (269, 0.47), big data (154, 0.22), data literacy education (28, 0.0), research data service (17, 0), research data (58, 0.16), open access (19, 0.09), research data management (38, 0.09) and so forth, and its individual contour value is 0.674. Through impactful analysis of the hot literature on data literacy research, the author has found that the studies from the library's perspective are mainly focused on data literacy education, which consists of three modules: training data consciousness, cultivating data ability, and establishing data ethics. It is urgently needed by the universities and the society at present and is especially important for researchers and a necessary condition for librarians to participate in research data management and service. Investigations have shown that different groups have significant differences in data literacy. As for university libraries, it is advisable to set data librarian positions, build data service webpages and develop diversified data literacy education to improve the data literacy of high school students and researchers and to effectively improve the efficiency of research data management by improving data literacy of researchers and librarians (Long, 2015). Today, society has entered the era of big data and “Internet +”, where information resources are becoming more and more abundant, and higher demand has been posed for data processing capabilities. In the future, library data literacy education can be achieved by traditional literacy education methods, such as training, publicity lectures, and supplemented by library + service practice and data management (Yang, 2015). The research direction at present and even in the future is to improve data literacy by reasonable educational methods and models and practices.

Data integration is defined to collect, sort, clarify and integrate the research data from different sources to form a new data source. The joint cataloging is mainly to adopt modern technical concepts to integrate and utilize the number and human resources of the libraries at all levels, realize joint construction and sharing of bibliographic resources, and avoid redundant construction of resources. It is one of the important ways of libraries' research data management and an important embodiment of data integration. The #4 data integration mainly includes 12 keywords, such as data integration (26, 0.05), information resource (23, 0.12), text integration (2, 0.01), multi-college integration (5, 0.02), university merger (2, 0.05), etc. The #3 joint cataloging mainly includes 14 keywords, such as library (253, 0.62), data sharing (34, 0.16), data reference (5, 0.0), catalog sharing (2, 0.03), data conversion (2, 0.03), data integration (2, 0.0), etc. For system transformation of the library or the merger of colleges, the university library needs to implement unified planning and management of the literature resources. Lu (2003) studied the integration of the library bibliography of New Jianghan University and pointed out that there are problems such as redundancy, waste of human resources, the difficulty of sharing, and low efficiency in the research data before integration. Zhu and Wang (2010) pointed out that the use of cataloging data, data integration of online cataloging organizations, sharing of trans-sectoral cataloging data, and social participation in cataloging mode might effectively promote the explosive growth of libraries' joint cataloging of data resources. The trans-sectoral cataloging and sharing would be an inevitable trend of future development of library cataloging (Liu and Zhou, 2011).

The development of libraries should provide strong support for developing scientific research and continuously optimize the service process to provide researchers with professional research support service. The #6 research support mainly includes ten keywords, such as data monitoring (36, 0.06), research support (10, 0.00), scientific research service (7, 0.07), data librarian (22, 0.01), digital learning (7, 0.00), library service (11, 0.02). The library is the link between the published literature and the research data. It can provide many cross-borders, embedded, and dynamic services for the data support of e-science and e-research, which also lays the foundation for the library to find a foothold in the new era. The research support service has also been one of the requirements for libraries to deepen services in the context of big data in recent years (Si and Zeng, 2018). The research support service of university libraries is mainly embodied in research data management, open access, academic publishing, research influence measurement, research navigation, research consultation, research tool recommendation. By studying scientific support of American libraries, the scholar Liu and Chen (2018) pointed out that the advanced experience of data librarian training should be learnt from American libraries and continuously improve the career development system of library data librarians in China. Xia et al. (2017) studied scientific research support of the libraries in 40 universities at home and abroad and found through the investigation that foreign research support services had relatively straightforward settings and services and comprehensive contents and were suitable for reference of the libraries at home. The scientific research support of the libraries should be improved in deepening knowledge and service levels (Song et al., 2017).

CiteSpace V software was adopted to visually analyze the time-zone distribution of keywords to dissect the evolution path of hot scientific data research topics. With the constant time division, the author chose and set the node as keyword, the threshold item as Top 20, and the output as “Time Zone” to obtain the evolution path graph of hot scientific data topics (Figure 5). In Figure 5, a series of keywords in each time zone represents a hot study topic in this time zone. The study of library scientific data began in the late 1980s and the beginning period was from 1988 to 1995, when related research was very scarce; from 1996 to 2000, the studies mainly focused on the sharing of library research data; from 2001 to 2003, college libraries and digital libraries entered people's field of vision, which was also the development of data sharing research in previous years and made data sharing and use better; from 2004 to 2006, studies mainly focused on scientific data management; from 2007 to 2011, the application of cloud computing became a major research topic; from 2012 to 2019, higher and more extensive demands were placed on the original scientific data due to the arrival of the data-intensive era in 2012, and in this period, impactful studies were conducted on data supervision, scientific data management, data service, and knowledge service. The new era has provided researchers with good data resources and research tools and put forward higher requirements for researchers' data literacy. Therefore, data literacy and its education were mainly studied during 2016–2019.

The Burst Detection algorithm was proposed by Kleinberg (2002). It refers to the sudden increase number in a short time, which has an intelligence function and can reveal the frontal aspects of research in this field. Figure 6 is based on the burst detection of keyword click. In this figure, Keywords are the corresponding ones, Year is the time when the search record appears for the first time, strength is the intensity of the burstiness, the beginning indicates the time to become the frontal topic, the ending is the closing time, and the start and end times correspond to one red rectangular block for a year. It can be seen from Figure 6 that the burst keywords up to 2019 are data literacy and its education, of which the corresponding burst strengths are 9.222 and 7.2756%, respectively. The data literacy research has been sustained for four years, while the data literacy education has only lasted for three years. It is a general trend for strengthening data literacy education in the era of big data, and the university library gradually transforms from traditional information literacy education to data literacy education (Zhang, 2018). The development and practice of data literacy education are currently in their infancy. The studies of relevant scholars in China are based on foreign university libraries, such as research course, research team and learning process of the data literacy education practice in Purdue University Library (Xu and Gao, 2018), the New England Data Management Collaborative Program of the University of Massachusetts, the Data Information Literacy Program jointly developed by Purdue University, University of Minnesota, University of Oregon and Cornell University, and the MANTRA Education Program at the University of Edinburgh Library (Hu and Wu, 2016). Some scholars have also researched current situation in China, but they have not carried out effective practice. Therefore, the impactful study on data literacy and its education practice and application is one of the leading research directions in the future.

The study aimed to review the literature on the topic of research data systematically. The researchers selected 1,238 studies fulfilling the inclusion criteria of the study. The reviewed literature revealed that research data is in an immature stage. Comparatively, research on research data is better observed in academic universities than other research institutions. We discuss the data analysis results as follows: (1) The publications of relevant literature in research data are increasing year by year, which indicates that the field has gradually gained attention from academic institutions. With the development of E-science and data-intensive research, theoretical and practical research in the field of research data will grow, and hot topics will change at different stages, (2) the researchers focus through three stages: the construction of research data sharing platform, data management and service, and data literacy education. Most of the literature in the first phase concerns practical cases of research data sharing. It includes four aspects: research data sharing platform construction, heterogeneous data resource integration and access (Liu and Zhu, 2007), the exploration of the influencing factors of sharing platform construction (Huang et al., 2008; Si et al., 2014), and the strategy of promoting research data sharing (Huang et al., 2014). The second stage is that scholars began to pay attention to the management and service of research data. These include research data monitoring service and library role adjustment (Zhu, 2014), big data and research data management, foreign scientific data management and service practice (Zuo and Chen, 2014), and joint cataloging of research data and library literature resources (Zhu and Wang, 2010; Liu and Zhou, 2011). In the third stage, many scholars pay attention to library data literacy education and training (Long, 2015; Yang, 2015). Data monitoring training and transformation of Information literacy education to data literacy education become the main issues considered by librarians and stakeholders, (3) the initial researchers were mainly stakeholders in the study of data. They explored the theory and practice from their respective disciplines and studied the construction of data sharing platforms and data management in their respective fields. Then librarians gradually participate in the research of research data. They use the research data as an information resource and study it from organization management and information service.

This study is a systematic literature review, and it is possible that some relevant studies might have been missed. Further, the data were limited to published studies between 1998 and 2018 and further limited to specific databases and sources. This paper uses quantitative research methods to analyze the relevant literature in the field of research data; hence, more studies using the mixed-method approach may be needed to understand the research data in depth.

Research topics gradually shift from technology to management, service, and policy from the path evolution of research hotspots. However, in addition to the issue, the policy documents such as “promulgation of the Measures for the Management of Research Data” and “Interim Measures for the Sharing of Government Information Resources”, the professional associations, research institutions, universities, and local governments should also formulate supporting implementation policies. Researchers need to continue to study foreign research data management policies, summarize the experience and improve the connection between macro and micro policies combined with the actual situation, and further promote the formulation and implementation of research data management and sharing policies in Chinese libraries.

Academic libraries have studied the theory and practice of research data management and service since 2011, but they are not in the core position in academia. For promoting academic libraries' research on research data, we put forward three practical implications: (1) Chinese higher education commission or ministry, funding agency, higher education institution and/or research commission should allocate budget to train researchers and librarians to raise their awareness and technical level of research data management services; focus on training librarians data management planning, data processing and analysis, data description, data sharing platform construction, data quality education, (2) academic libraries and stakeholders should pay attention to the transformation of research hotspots in the field of research data and promote the development of library information services, (3) academic libraries should use collaborative governance theory, data life cycle theory, stakeholder theory, data asset theory, research data technology to construct library research data governance model, research data management and structure system. In this way, the relevant theories and technologies of the research data will be combined with innovation.

Because the research data field is still in its infancy in developing countries such as China, the research data literacy and data management awareness of researchers and stakeholders are relatively weak, so the collaborative relationship between scholars is loose and needs to be strengthened urgently. Higher education institutions and/or research boards, funding agencies and higher education commissions or ministries should sit together and make it compulsory while granting funding to researchers to submit their research data in their institutional or subject repositories and publish their work in open access journals. The research literature of academic libraries around research data management and service will continue to increase in the future. They should cooperate with multidisciplinary researchers to carry out interdisciplinary research data management and service research to expand the boundary of research data.

From the perspective of library, this paper establishes a visual knowledge map for the subject literature of Chinese research data by using bibliometrics and social network analysis methods and focuses on the volume and collaboration of authors on the subject in the past 30 years, the volume and collaboration of scientific research institutions, the research status and future research trend of library research data. It can be used as a reference direction for scholars to study all aspects of research data with the library as the starting point. The results show that (1) among the related research on the subject of research data in the library, Gu Liping and Shen Tingting have the most papers, but the collaborative relationship between scholars is loose and needs to be strengthened urgently; (2) the scientific research output institutions of library research data subject research are in the top three: Wuhan University, Shanghai University Library and document Information Center of Chinese Academy of Sciences, and the collaborative relationship between scientific research institutions is not close, and it is necessary to strengthen collaboration; (3) from the cluster analysis of keywords, we get the research hotspots related to the research data, they are the hotspots of library science subject research include the construction of research data sharing platform, librarian data literacy education, research data integration management and library joint cataloging, library research data service mode, library participation in research incentive mechanism and policy support, library to verify the effectiveness of research data; (4) from the path evolution of research hotspots, we know that the theme of research data has experienced three stages of data sharing platform construction, data management and service, and data literacy education, (5) burst detection tells us that in-depth research of data literacy and practical application research of data literacy education are the main research directions in the future.

Through the in-depth interpretation of the related papers about research data, this paper puts forward some suggestions on the future research direction: (1) In the era of big data, cloud computing technology has been gradually applied to the research data sharing platform because of its high security, various data types, fast access, and low energy consumption, which further improves the sharing ability of research data and ensuring its security. In the new era, the data sharing mode of the library will be fully developed, and the construction of the unified data sharing mode, the convenience of data functions on the sharing platform and the improvement of data analysis capabilities will be essential research contents in the future. (2) Based on profoundly analyzing the connotation of data literacy, research data workers should combine the national conditions of China and the future library development plan and deeply study how to adopt effective and operable practical strategies or solutions to embed library data literacy education in every link of other courses or activities. Scholars should study the technical difficulties faced by data literacy work, librarians' ability requirements, and integrating with other library services and education. (3) Joint cataloging is an important and effective way to effectively manage, share, integrate, and develop library data and solve traditional library data problems. Therefore, it is required to continuously improve the joint cataloging of libraries and the trans-sectoral, trans-class, and trans-level cataloging mode to better integrate, manage and share library data resources. (4) The librarians should provide better support services for scientific research. It is necessary to continuously improve the data structure and cultivate the professional quality of the subject librarians and deepen service levels (e.g. fully participating in scientific research, promoting the publication of research data, standardizing data references, and providing research data quality assessment) and carry out impactful studies in other aspects. In addition, scholars should increase systematic empirical research and strengthen interdisciplinary collaboration to find out the unique value of library research data through joint exploration of research data with other disciplines. In the research direction, scholars should carry out basic research based on library and information discipline characteristics, such as the formulation of scientific data management policy, the professional literacy of relevant personnel, and the intellectual property rights of scientific data.

This study was substantially supported by a project (70901080) from the National Natural Science Foundation of China. The research was substantially supported by a project (19SKGH091) from the Chongqing Education Commission of China and project from the Open Fund of Research Centre of Enterprise Management. This study was also substantially supported by a project (2053002) from Chongqing Technology and Business University of China.