Automatic occupation classification based on the KSA model for Spanish job texts in the Mexican automotive sector

Abstract

This thesis explores how competency information—specifically Knowledge, Skills, and Abilities (KSA)—can support the classification of Spanish job postings into the Mexican occupational taxonomy SINCO 2019. While international frameworks such as ISCO and ESCO have guided occupational modeling for many years, there is still no practical method that connects KSA evidence in Spanish job texts to SINCO categories. This gap is especially visible in Mexico, where job postings vary widely in structure and vocabulary. To address this challenge, the thesis develops a transparent classification pipeline based on TF-IDF representations of normalized KSA terms combined with simplified job-title information. The study evaluates different feature configurations using reproducible experiments and shows that combining KSAs with title cues leads to more stable and interpretable predictions than using titles alone. The results highlight which SINCO unit groups are harder to distinguish and how overlapping competencies influence misclassifications. Overall, the thesis provides a practical baseline for Spanish occupational classification and opens opportunities for future work, including hierarchical approaches, automatic KSA extraction, and transformer-based models adapted to Mexican labor data.