Interpretable Job Recommendation from Multiple Sources of Occupational Evidence: Early Results

• Eric Keiji Tokuda Universidade de São Paulo (USP)
• Alexandre Cláudio Botazzo Delbem Universidade de São Paulo (USP)
• Anderson da Silva Soares Universidade Federal de Goiás (UFG)

DOI: https://doi.org/10.5753/sbbd_estendido.2026.249748

Resumo

Skill-based occupation recommenders should show how evidence from different data sources leads to ranked occupations, especially for job seekers with limited literacy or access to guidance. This paper presents a deterministic pipeline for traceable job recommendation that integrates structured occupational descriptors with annotated job-text spans and converts them into yes/no questionnaire evidence. The pipeline builds O*NET occupation profiles, audits SkillSpan skill and knowledge spans, maps them to occupational descriptors, generates traceable questions, links questions back to occupations, and tests whether simulated answers recover the source occupation. The question structure works well under controlled reconstruction: compact task-based questions recover occupations almost perfectly. The multi-source text-integration branch is weaker: lexical extraction predicts 9,375 spans for 2,264 gold spans, and conservative normalization accepts only 662 mappings. The pipeline is therefore a job-recommendation prototype and audit layer for inclusive occupational evidence, showing where evidence remains usable and where extraction or mapping breaks down.

Palavras-chave: text-integration, data fusion, recommendation

Referências

Converse, P. D., Oswald, F. L., Gillespie, M. A., Field, K. A., and Bizot, E. B. (2004). Matching individuals to occupations using abilities and the o* net: Issues and an application in career guidance. Personnel Psychology, 57(2):451–487.

Gavrilescu, M., Leon, F., and Minea, A.-A. (2025). Techniques for transversal skill classification and relevant keyword extraction from job advertisements. Information, 16(3):167.

Ghosh, S. et al. (2023). JobRecoGPT: LLM-based job recommendation interfaces. arXiv preprint.

Ko, H., Lee, S., Park, Y., and Choi, A. (2022). A survey of recommendation systems: recommendation models, techniques, and application fields. Electronics, 11(1):141.

Lawler, E. and Ledford, G. (1992). A skill-based approach to human resource management. European Management Journal, 10(4):383–391.

Zhang, Y. et al. (2020). Explainable recommendation: a survey and new perspectives. arXiv preprint.