期刊名称：Eastern-European Journal of Enterprise Technologies
印刷版ISSN：1729-3774
电子版ISSN：1729-4061
出版年度：2017
卷号：5
期号：2
页码：4-11
DOI：10.15587/1729-4061.2017.112323
语种：English
出版社：PC Technology Center
摘要：A method for the clustering of scientific publications is proposed in order to identify areas of scientists' research areas. In this method, the links between scientific publications and citations are represented in the form of a directed graph. There are two proposed techniques for finding a distance between publications in the method for clustering the scientific publications. The first technique is based on the calculation of the length of the minimal route between the corresponding vertices of the graph of links between publications through citation. The second procedure is based on the calculation of the degree of closeness by the content of abstracts of these publications using the Hamming distance on the basis of a locally-sensitive hashing method. After the application of the method for clustering this graph, considering the specificity of input data, it is proposed to merge clusters by the criterion of proximity of centers of gravity.To identify scientists' research areas, it is proposed to initially use one of the expert methods for establishing a correspondence between the built clusters and the appropriate verbal representations of scientific areas. Next, to form for each scientist a set of areas for scientific research, taking into account the mapping of a set of scientists onto a number of scientific areas.The methods proposed could be used in scientific and educational institutions, as well as private companies that are engaged in the creation of science-intensive technologies.
关键词：clustering;area of scientific research;co-citation graph;locally-sensitive hashing
