Thematic and paradigm models of the concept system of science (a case study of advanced research in Russian linguistics)

Abstract

The article describes two approaches to modeling the concept system of science - the thematic and paradigm ones. The research represents a case study of the two corpuses of abstracts: abstracts of projects supported by the Department of Humanities and Social Sciences of the Russian Federal Property Fund in linguistics, as well abstracts of articles by authors (and their co-authors) who have received multiple support from this foundation. Thematic modeling was carried out within the frameworks of two approaches: сcorpus based approach (modeling the system of concepts as a holistic entity of term fields network), and single text approach (singling out compositions of term fields steadily present in the texts of projects which can be treated as separate branches of linguistics). The method of semantic graph modeling realized in the "Semograph" Information system (corpus based approach) and the K-means clustering method (single text based approach) were applied. Arrays of keywords (those that occurred in the abstracts of projects supported by the Foundation) grouped into term fields served as operational units. Paradigmatic modeling was based on the analysis of network interaction of researchers (created on the basis of the information on joint publications). After the hypergraph of researchers (2,108 state points) had been created, it was divided into subgraphs (network communities) by means of the modularity method (analogue of cluster analysis). Each cluster can be considered as a model of a scientific community that actualize a certain scientific paradigm; a set of clusters represents a model of the concept system of a certain subject domain from the point of view of representation and interrelation of several paradigms. The synthesis of thematic and paradigmatic models appears to be the direction for future research that involves modeling the concept system of science. The considered models can be applied as an effective means of monitoring, forecasting and managing scientific research, i.e. as an instrument of state and/or department policy.