Bridging psychometrics and language: a method for extracting leadership insights from open-text responses

We present a compact, reproducible NLP method that turns open-text leadership feedback into theory-aligned signals and validates them against questionnaire scores. Inputs are multilingual 360° feedback. After preprocessing and translation, we (i) classify sentiment and (ii) compute construct salience scores by calculating cosine similarity between embedding space open-text feedback and seed-phrase representations of Deep Leadership Model (DLM) constructs. We test the estimated scores and classes against validated questionnaire results using three criteria: (1) association between sentiment and overall questionnaire outcomes with controls for open-text feedback type ; (2) construct salience score correlation with matching questionnaire factor scores versus non-matching and permutation baselines; and (3) interpretability via 360° role-wise construct profiles that align with established patterns. Results show that framework-aware open-text scoring complements existing DLM metrics and provide transparent, auditable diagnostics at the construct level. Because the approach relies on seedable constructs and questionnaire anchors, it generalizes beyond DLM: the same pipeline can augment any psychometric tool that pairs open-text responses with theory-defined dimensions, supporting scalable development, monitoring, and evidence-based use.