A Metadata Application Profile to Structure a Scientific Database for Social Network Analysis (SNA)

There are a number of challenges associated to metadata in its different applications including data quality, data acquisition, computing resources, interoperability, and discoverability. This work presents an approach to structure metadata of scientific information for social network analysis based on an academic case study from scientific articles published by universities, to evaluate the area of risk assessment. Studying metadata for scientists' social networks helps identify authors' relevance based on their position within the network. By using Elasticsearch (ES) and Python technologies, this work addresses big data analysis issues related to data structure and volume, given ES full-text search engine capabilities for indexing and searching data, and Python's processing support. The data is obtained from the ArnetMiner (Aminer) open scientific database providing a fresh overview of scientific records up to January 2019. From a sample of 64,070 publications, a total of 45, 000 relations are graphed in a co-authorship network. Through the computation of network centrality measures, this work identifies central-positioned authors, clusters of research, and their affiliations. The results show that degree centrality is an important measure to identify prominent scientists in this co-authorship network, and closeness and betweenness centralities together are dominant measures to pinpoint the key players in the flow of information within the network. We conclude that the application of this approach allows rapid full-text search, visualizing dense co-authorship networks, and identifying central authors through centrality metrics. The results presented in this work can help researchers or research groups identify key research collaborators, multi-disciplinary areas, and international stakeholders.