Dobrovolskyi H. Model, Method and Information Technology of Scientific Publication Selection in the Process of Bibliographical Index Compilation

The object of research is the process of preparation of scientific-supporting and recommendation bibliographic indexes in scientometric search systems. The objective of the research is to increase the efficiency of discovery and selection of scientific publications in the process of preparation of scientific-supporting or recommendation bibliographic index by obtaining the minimal terminologically saturated ordered set of the publications with the employment of the engineered information technology that is based on the developed model, method and software implementation. Research methods are the means of the set theory and methods of graph analysis to describe the developed models and methods; machine learning and natural language processing algorithms for topic modeling and automatic detection of terms; methods of graph analysis for detection and selection of the most influential publications, mathematical statistics for calculation of criteria for stopping bibliographic detection and selection. Results are: the hybrid mathematical model of the process of bibliographic detection and selection, which differs from the analogs with simultaneous use of iterations of controlled snowball sampling, probabilistic topic modeling of text documents, analysis of the citation network, automatic detection of terms and terms; a method of bibliographic detection and selection that uses the model; information technology that uses the model and method; software implementation of the information technology; the improved with principal component analysis method of probabilistic topic modeling of text documents; advance in information search methods designed to meet the information needs with low specificity. Introduced in the educational process of the Ukrainian Catholic University and Zaporizhzhya National University of the Ministry of Education and Science of Ukraine.