Insider Risk Evaluation Based on Sentimental Analysis of Digital Communications

The paper presents the development and implementation of methodology for insider risk evaluation based on sentimental analysis of digital communications within Customer Relationship Management (CRM) systems, using the Fitbase platform as a case study. The proposed system architecture includes modules for text and speech sentiment analysis, a centralized metric storage layer, and a visualization and alerting component. Text analysis is carried out using transformer-based models (e.g., RuBERT) trained on Russianlanguage corpora. A generalized monitoring framework is introduced, comprising stages of data acquisition, sentiment normalization, emotional vector tracking, and risk index computation. The approach is positioned within the domains of information security, data loss prevention, human resource analytics, and behavioral anomaly detection.