An unstructured data-driven methodology for knowledge extraction on a complex network of scientific productions
Bibliometrics, Scientometrics, Machine Learning, Complex Network Analysis
Science plays a fundamental role in the world, because it enables the social and economic development of a country. Due to its importance, it is essential to analyze and measure it, using the techniques of Bibliometrics and Scientometrics, to understand how knowledge is produced over time, between disciplines, around the world. In addition, these types of studies also allow identifing the trends and productivity in science and technology, being essential subsidies for creating or adapting research promotion policies, with the purpose of expanding the scope and potential of science, from the required financial and institutional resources. However, due to the exponential growth of the scientific literature, a systematic approach is necessary to enable monitoring and an overview of scientific knowledge that is continually expanding. Given this context, this study aims to improve the process of bibliometric and scientometric analysis, through a methodology based on Machine Learning and Complex Network Analysis. The proposed methodology will apply Machine Learning techniques in complex networks of scientific productions, as well as it will be validated experimentally. The preliminary results of two case studies indicated the feasibility of the proposed solution in mapping the dynamics of science, as well as they demonstrated the potential for decision-making, either in evaluations of scientific entities, or for promoting research.