Bibliometric analysis of document ow on academic social networks in Web of Science

Analysis of a document array on academic social networks (ASNs) in Web of Science for the period from 2005 to 2020 was carried out with use of analytical services data of the WoS and CiteSpace (the program for visualization of patterns and trends in scientic literature). The following parameters of the array were analyzed: publication dynamics; document types structure; countries, organizations and authors leading in the number of publications; thematic categories to which documents of the array are assigned; publications (journals, monographs) in which the documents of the array are published; most cited publications. An increase in the number of publications on the ASNs in WoS was established since 2005. The largest number of ASNs studies is conducted in the USA (University of Pittsburgh), UK (Wolverhampton University, Manchester University), China, Spain (University of Granada), Germany (Max Planck Society for Scientic Research), Canada, India and the Netherlands (Leiden University). The ASNs were studied in the main thematic areas: Computer Science, Computer Science and Librarianship, Mechanical Engineering, Engineering and Technology. Four out of the rst ten highly cited publications, are devoted to altmetrics. Using the CiteSpace, it was shown that when ASNs started rise, their organizational structure was beeing studed. Later, altmetrics used in the ASNs became the main subjects of ASNs research. The keywords occurrence revealed that the most frequent terms are “altmetrics”, “impact”, “citation”. As part of the document ow, also identied publications in which the ASNs are used as a source of bibliographic data for systematic or meta-analysis (in medicine predominantly), or as a platform for experimental data discussion.

When searching, we faced the problem of separating publications: in some, academic social networks were mentioned as a source of bibliographic information for simple review, systematic review, metanalysis, in others -as a place to store data, in still others -as a place to discuss certain professional issues. We were interested in publications where academic social networks were the object of scienti c research. We tried to exclude such a document by using the not operator and listing addresses (for example, "not TS = (https://data.mendeley.com OR" Mendeley Data "OR" https://www.researchgate.net "OR myexperiment.org ), but this did not help us. When browsing the selection with exceptions using the NOT operator, we continued to nd publications of the above nature, for example, publications where the ResearchGate network was simply listed as a source of bibliographic information. For this reason, we analyzed the entire document ow.

Methods
WoS. Analytical services of the WoS database were used to obtain information about the investigated document array by the following parameters: dynamics of publications by year; type-speci c structure of an array of documents; leading authors in terms of the number of publications; leading organizations in terms of the number of publications; thematic categories to which the documents of the array are assigned, the distribution of publications by country; editions (magazines, monographs) in which the documents of the array are published; the most frequently cited publications.
CiteSpace. The CiteSpace, program for visualizing patterns and trends in scienti c literature, used to analyze the co-citation of documents, co-occurrence of keywords in documents, and names of institutions.
Documents co-citation analysis is one of the ways to determine the relationship between various publications in a certain eld of research: if two publications are cited in the third jointly, then with a high degree of probability they belong to the same research area (Small 1973). Clustering of the document citation network and automatic tagging of clusters with the author's keywords, additional WoS keywords of citing articles or terms from the titles of citing articles or terms from abstracts using the CiteSpace program allow you to identify research trends in the analyzed document array.
Collaboration networks are plotted in CiteSpace on the analysis of the co-occurrence in the document of the original array (1216 documents from WoS, in our case) of the scienti c institution' names (institutions collaboration) or authors' names (authors collaboration).
The analysis of the co-occurrence of keywords in the titles of articles, abstracts, among the author's keywords and Keywords Plus WoS of the original documentary array on the ASNs topic in CiteSpace allow us to identify their frequency distribution and thematic clusters in which they are grouped.
Clustering the network of co-occurred keywords and labeling them with terms from the titles of articles in which they occur together allow us to identify the topics of the investigated document array.
The results of the analysis of co-citation of documents, co-occurrence, institution names (author names), keywords are presented in the form of pictures of the corresponding networks and clusters, in which the corresponding analyzed parameters (references, authors, organizations, keywords) are grouped.

Results And Discussion
Document types and publication languages According to analytical services WoS, the bulk of documents are scienti c articles in journals (54.4% of documents), in conference proceedings (23.8% of publications) and reviews (18.2%) ( Table 1). About 95% of the documents found in WoS are published in English, which is logical for the international scientometric base.    Most frequently cited publications on the topic of "academic social networks" The ten most frequently cited articles (Table 8) on the ASN were published from 2009 to 2015. in such publications as "Nucleic acids research", "Scientometrics", "Proceedings of the national academy of sciences of the United States of America", "Journal of Informetrics", "Journal of the association for information science and technology". On average, such articles are cited 11 to 38 times per year. The total number of citations varies from 105 to 305.
The total number of citations of publications in the collection in July 2020 is 5813, the average number of citations of the document is 6.8, and the Hirsch index is 35.
Among the most frequently cited publications is the work of Wolstencroft K. et al. 2013 (Wolstencroft et al. 2013) performed at the School of Computer Science, University of Manchester. The article by Wolstencroft K. et al., 2013 is devoted to the Taverna platform, which hosts tools for processing biological research data. This article appeared as part of the scienti c social media work ow due to the fact that the repository for the working materials of these studies is the NCC myExperiment. A frequently cited article by Goble C.A. is dedicated to the same network. et al. (Goble et al. 2010).
Noteworthy is the fact that out of dozens of highly cited publications, six are devoted to academic reputation on-line and altmetrics.  Table 8), but by their presence, mention and use on the Internet and traditional media (social weight, social impact -highlighted in blue in Table 8). In the ASN, such altmetrics are actively included. In Two articles from the list of actively cited (Thelwall and Kousha, 2014;2015) are devoted to scienti c social networks ResearchGate and Academia.edu, respectively (Thelwall andKousha 2014, 2015), Thus, the top 10 highly cited articles indicate that scienti c social networks are of interest to the scienti c community as tools for assessing academic activity on the Internet, as objects for analyzing their structure, as repositories for placing scienti c data, as objects for developing algorithms for analyzing scienti c social networks. (computer modeling of processes, including). In total, the studies in which the ASNs are featured are classi ed in 118 thematic areas. The most quantitatively lled directions are re ected in Table 9.

Mapping the ASNs document array based on the document co-citation analysis
The document co-citation network (Fig. 3) in CiteSpace is characterized by the number of nodes and links of citation. Each node in the document co-citation network is a separate reference in the ASN document array in WoS. 662 nodes were identi ed, networked by 2943 co-citation links. The color of the link is determined by the year the document was rst co-cited. The size of a node is determined by the number of co-citations of a given document. Table 10 presents the rst 11 most frequently cited articles. In Figure   2, they are indicated by the name of the rst author and the year of publication of the rst 11 (out of 666) documents with high co-citation values (Table 11). Table 10 References ranked by the number of citations in the ASN document co-citation network (highlighted in Fig. 3)

Reference cocitation
Year of publication User, disciplinary structure of ASN, other aspects of functioning, comparative analysis of ASNs (Bhardwaj 2017) DCA revealed 112 clusters 8 of which are visually presented on the Fig. 4, 5. For automatic marking of clusters (Figs. 3 and 4), terms from the titles of citing documents (citers) were selected by the CiteSpace program and ranked on the bases of the algorithm of the algorithm using LLR (log-likelihood ratio test) (Chen et al. 2010). In Fig. 4, 5 clusters are labeled with the same term, in tab. 12 -three or four (CiteSpace identi es 100 terms, ranked in descending order of value of the applied algorithm (LLR)). Table 12 Articles in the document co-citation network with betweenness centrality value exceeding 0.1.

Cocitation
Betweenness centrality Publication year At the initial stages of the formation of scienti c social networks, the problem of content systematization was solved (cluster 5). Social network developers have followed a path that has already been tested in the creation of document repositories and digital libraries: labeling content with descriptors (descriptive terms), also called keywords or tags. You can navigate, lter or search by tags. In the case of a repository and a digital library, tagging was carried out by the creators of the resource (repository, digital library) using the data provided by the author of the document, and in the case of social networks, the content indexing function was provided to users and took the form of social tagging (collaborative tagging), which is also called Volksonomy (folksonomy = folk taxonomy) (Hotho et al. 2006;Rawashdeh et al. 2013). In the article, which is the main citer of the cluser (citing 10 out of 25 publications of the cluster), scientists perform a comparative analysis of thematic indexing of scienti c articles using the social tagging method (CiteULike) and the professional indexing method (PubMed).
In article (Kipp 2011) , which is the main citer of the cluser (citing 10 out of 25 publications of the cluster), scientists perform a comparative analysis of thematic indexing of scienti c articles by the social tagging method (CiteULike) and the professional indexing method (PubMed).
Later (cluster 3, 2007) the problem of organizing services by recommending content to users based on social tags, collaborative tagging (Community recommendations), appears.
Cluster article Golder S.A., Huberman B.A. (2006) (Golder and Huberman 2006), which is highlighted in the network as the key one (BC = 0.10) ( The next in chronological order is cluster 28, which highlights the analysis of full text sources in the Google Scholar search engine as a separate topic. Of the individual sites, ASN ResearchGate is the most productive source of full text in this system (Rohani et al. 2014).
Cluster 14 (average year of publication of cluster documents -2010) denotes research on the problem of developing recommendation systems for ASNs. Recommender systems are programs that are designed to predict which objects (movies, music, books, news, websites) will be of interest to the user, given certain information about his pro le. When creating recommendation systems, the "cold-start problem" comes rst. The "cold start problem" is a problem solved in the development of recommender systems in social networks. The problem is what the ASN should offer to a new user of the network and to whom to offer new content (Rohani et al. 2014).
Further, in terms of chronology and localization in the network, there are clusters (2, 0 and 4) with various aspects of studying the altmetrics "Social science", "Case study", "Mendeley reader" (Table 11). It is no coincidence that among the cited articles on the "path to these clusters" is the publication "Manifaesta Technology Facilities Council)) made a proposal using alternative metrics (alt-metrics, altmetrics) [4]. The essence of their proposals is expressed in the Altmetry Manifesto (J. . Almost the most numerous cluster 1 symbolizes the research front that studies various aspects of ASN: user composition, disciplinary structure of networks, comparative ASN research (Table 11).
The fact that alternative metrics of the ASN and the assessment of scienti c impact on their basis are actively studied in the scienti c literature is con rmed by our research using the CiteSpace program for the co-occurrence of keywords in titles, abstracts, indexed terms of articles from the array of documents on the ASN. The terms most frequently encountered in this array are "altmetrics", "impact", "citation" (Table 13).  The results of cluster analysis of the network of co-occurrence of keywords were interesting (Fig. 5). In this case, thematic areas were revealed that were not found in DCA. These directions are not connected with the direct study of the ASNs, but with their use as a source of bibliography for performing a review, systematic review or meta-analysis clusters 11,2,29,13 (Wolstencroft et al. 2013). This tool came into our eld of vision for the reason that it is compatible with the repository of work ows (work ows repositories) myExperiments, which positions itself as a scienti c social network social network.
All these clusters, reviled using co-word analysis, are not identi ed using document citation analysis, since they are not united by a common research problem and, therefore, a common intellectual base in the form of a set of cited publications, but the use of social networks as a source of literature or a repository of work ows (work ows).
It should be noted that the research topics of such authors as Abramov, Valery M. and Gogoberidze, George G. (Russian State Hydrometeorological University, Saint Petersburg, Russia) are not explicitly re ected in the research fronts identi ed by CiteSpace. Moreover, both authors are included in the list of authors with a large number of publications in the studied array of documents (Table 7). The studies of these scientists are classi ed in WoS as Geosciences Multidisciplinary; Water Resources; Computer Science Interdisciplinary Applications, etc. The works of these scientists were included in our array for the reason that abstracts of all publications of these authors contain a link (https://www.researchgate.net/pro le/Valery_Abramov2/) to V. Abramov's pro le in Researchgate. "The platform gave excellent opportunities to preliminary discussion and data exchange in the frame of these researches", writes V. Abramov. Thus, in this study, these works form the research front of cluster 0 (Table  14).
[2] In February 2019, CiteULike announced that it would be ceasing operations as of March 30, 2019).
[3] Connotea discontinued service on March 12, 2013) [4] Altmetrics is a discipline whose subject is the creation and research of new metrics (alternative metrics) for evaluating a scienti c product (articles, books, presentations, statements and discussions on the topic of scienti c research, computer programs, etc.) within the virtual space (the number discussions on social networks, downloads and views in scienti c repositories and bibliographic managers, etc.).

Conclusions
Academic social networks are a new type of scienti c communication that is implemented in the electronic environment. Following the formation of global social networks based on Web 2.0 technologies, specialized academic social networks were developed to meet the needs of scientists. As a phenomenon of social communication, global social and scienti c social networks are the object of study of scientists. A large number of scienti c works are devoted to the study of global social networks.
For example, in Web of Science, only in response to the query TS = (Twitter OR Facebook), we found 34,150 documents (as of January 21, 2020). The scienti c document array for the ASNs topic is more modest in size. In this work, a multidimensional scientometric study of the document array presented in the WoS is carried out. Countries, organizations, scientists which are investigating or using the modern form of communication between scientists -academic social networks, are identi ed. It is shown that research on ASNs began to be conducted in 2005. Leaders in research on ASNs are scientists from Great Britain, Germany, USA, Spain. At present, the main issues discussed in connection with the ASNs are alternative metrics designed to re ect the activity of scientists on-line, developed for a number of networks. This is evidenced by the fact that among the most frequently cited documents according to the WoS database, there are many articles devoted to this topic. Using the program for visualizing patterns and trends in scienti c literature CiteSpace, a network of citation of documents for an array according to the ASNs was built, the clustering of which made it possible to identify the main areas of research from 2005 to 2019. social tagging and professional indexing, the problem of "cold access" was solved.
Subsequently, in some ASNs (Mendeley) systems for assessing scienti c activity in the web environment (altmetrics) were developed. In addition, the organization of various ASNs, the behavior of their members, etc. are studied. However, altmetrics in ASNs are the main research trend. In addition, in this study, the CiteSpace program was used to identify the nature of collaboration between organizations in the study of ASNs.
It should be noted that, in our sample from WoS, along with articles that directly analyze ASNs, there are articles in which ASNs are referred to as scienti c data repositories (Mendeley, myExperiment). Among them, there are articles with high citation rates (Goble et al. 2010;Wolstencroft et al. 2013). All this indicates that the document ow we are analyzing re ects not only the fact that ASNs are the object of scienti c research, but also the fact that ASNs have entered directly into the life of scientists.
Academic social networks continue to evolve. For example, there are messages about the termination of activities of some of them (CiteULike, Connotea). Time will show what the further ways of transforming scienti c social networks will be. As well as the developing phenomenon of scienti c communication, ASNs require further research.