Improvement of the method for scientific publications clustering based on n-gram analysis and fuzzy method for selecting research partners

Abstract

For the problem of formation of project teams, in particular, scientific research project groups, there was proposed the comprehensive method, which consists of the two-stage method for clustering the graph of citation of scientists» publications and the method of fuzzy inference for coordination of experts» opinions on the selection of potential partners and their inclusion in the project group. The essence of the two-stage method for clustering publications of scientists is clustering the citation graph based on the proximity of abstracts of publications. The distance between publications is calculated based on the determined metrics and approaches of the n-gram analysis. The described method allows identifying the areas research of scientists, which is a necessary component of the rational choice of a partner for the formation of a project team and is the input information for experts who form this group. The next step is the application of the method of fuzzy inference, which is constructed to coordinate opinions of experts on the creation of project teams. This method consists of three stages. At the first stage, fuzzification is performed through the introduction of function of scientist»s belonging to the area of scientific research. The second phase of fuzzy inference is the statement of experts» requirements to candidates for a place in a project group. At the final stage, defuzzification with the use of the method of the weight center takes place. To verify the fuzzy method for identification of research project groups, the organizations-executors for a fundamental scientific research were determined. Described methods can be used for the problem of formation of scientific research groups and identification the similarities between the fragments of text information based on the n-gram analysis, which is used in the problem of identification of incomplete duplicates between fragments of text information.