Methodology for determining the level of remuneration based on the analysis of professional competencies using large language models

The article presents a methodology for determining the level of remuneration based on a multifactorial analysis of professional competencies using language models. The relevance of the study is due to the growing differentiation in the level of remuneration for specialists within the same professional category and the subjectivity of existing competence assessment systems. The aim of the work is to develop an objective methodology for determining the level of remuneration based on a comprehensive analysis of textual vacancy data. The research methodology is based on the integration of qualitative (content analysis, semantic and contextual analysis) and quantitative methods (correlation and regression analysis) using natural language processing tools. The proposed methodology includes a three-step process: a preliminary analysis of explicit skill indicators, a contextual analysis of professional vocabulary, and the calculation of the final competency scores using weighting factors. Based on the results obtained, a graded remuneration system was formed, including 12 grades for four skill levels, and a regression model was developed to predict the optimal remuneration level. The results of the study revealed a pronounced correlation between the position grade and the average value of the salary range (r = 0.87), the required work experience and salary level (r = 0.79), as well as a significant regional differentiation of remuneration with a coefficient of 1.5–2.2 between the capital and regional centers. The approbation of the methodology showed an increase in the accuracy of forecasting the level of wages by 32 % compared with traditional approaches. The study was conducted as part of the research project Theoretical approaches and methodological support for labor market analysis in the Republic of Belarus using big data with the financial support of the Belarusian Republican Foundation for Basic Research under the G24-013 agreement..