Burnout is often conflated with depression and anxiety due to overlapping symptoms, underscoring the need for refined detection methods. Building on the conceptual framework of Maslach and Leiter (2016), this study integrates a custom-developed questionnaire with transformer-based Natural Language Processing (NLP) models to identify unique linguistic markers of burnout. Data collection has begun, focusing on professionals in high-stress settings who complete open-ended items that probe emotional exhaustion, cynicism, and perceived efficacy. Preliminary analyses, conducted on an initial subset of responses, suggest that participants endorsing burnout-related experiences more frequently employ language reflecting detachment and diminished fulfillment. However, overlaps with depressive expressions indicate that refining feature engineering remains essential.
In the first phase, questionnaire validity was assessed through expert review and pilot testing. Currently in the second phase, partial NLP model training has demonstrated promising early accuracy, but further data acquisition is in progress to bolster representativeness. Preliminary scoring of open-ended responses indicates consistent patterns of cynicism and frustration among high-burnout respondents, hinting at distinct lexical cues that may effectively separate burnout from anxiety or depression. Nevertheless, more extensive data gathering is required to conclusively validate these patterns across diverse occupational contexts. This integration of self-reported responses with advanced text analytics addresses the growing need for reliable, context-specific burnout measures and clarifies distinctions between closely related psychological conditions.
Findings from this study are anticipated to provide actionable insights for mental health professionals and organizational decision-makers, contributing to more targeted and effective approaches to mitigating burnout.