Unsupervised Co-Authorship Based Algorithm for Clustering of R&D Trends at Science and Technology Centers in Oil and Gas Industry

Planning of research and development trends in science and technology centers should be in line with the actual state of things. Such phenomena as organizational frigidity, research diversification and propensity for developing IT products are able to significantly impair any strategies and development trends. However, feasibility of plans is an important attribute of development able to which significantly raise personnels motivation for achieving best results. This is why setting achievable goals is of such importance. There are never enough quantitative tools for appraisal of research and development activities. Formal paperwork reporting on R&D is not suitable for evaluation of researchers involvement and dedication. Instead, small formats of research works such as presentations at scientific and technical conferences or scientific articles in peer-reviewed scientific publications require much more informal approach from researchers. Analysis of a science and technology centers performance based on its publication activity is a common practice. Many studies analyze text corpus of scientific articles and make conclusions on development trends. Text data noise levels are quite high; even most advanced analysis methods based on word embedding are able to produce accurate predictions only if analyzed are huge text volumes which are seldom available in case of small organizations. Small research organizations suffer the most from inaccurate planning of research activities. Authors of this research propose to take advantages of articles (presentations) analysis based on co-authorship bipartite graph to extract research trends with the purpose of their further evaluation and planning.