Development and Verification of a Digital System for Assessing the Linguistic Vector of a Linguistic Personality's Interpretation Potential (Based on Student Spoken Discourse)

The paper conceptualizes and verifies the linguistic vector as a core element of the author’s three-vector structure of a linguistic personality's interpretation potential. The study addresses the high subjectivity of traditional descriptive linguistic analysis, substantiating the mathematical description of complex cognitive-speech phenomena. It serves as a proof of concept for a universal "weighted" approach to modeling humanitarian reality. The research aims to develop, theoretically ground, and experimentally verify a comprehensive methodology for objectively measuring the parameters of the linguistic vector of human interpretation potential, integrating digital expert systems. This toolkit converts respondents' qualitative language abilities (semantic deduction, syntactic flexibility, metaphorical/distributive analysis, and paronym differentiation) into verifiable quantitative indicators via multi-level weighted data aggregation (micro-, meso-, and macro-levels). Artificial intelligence serves as an independent expert for the parametric evaluation of discourse characteristics. The methodology was tested on 16 students across five task blocks, with results compared against Unified State Exam (USE) scores. The scientific novelty lies in actualizing the linguistic vector by integrating linguistic analysis with multi-level data convolution and neural network expert systems. The experiment proves that macro-analyzing final scores is impractical due to multiple convolution effects. The study reveals a weak correlation between direct survey results and USE scores, which is attributed to differences in the measured constructs. However, a strong correlation (0.67) is found between USE scores and individual AI error forming at the micro-level. The paper demonstrates that the variability in AI ratings reflects the ambiguity of the responses themselves rather than algorithmic error. The study delivers a fundamental conclusion: systemic thinking and a high level of training among respondents (high USE scores) logically determine the stability of their cognitive and speech strategies, which in turn minimizes the individual error of the expert system.