Elamur Research⌕
☰
home / papers / 2026 / UniSkill: A Dataset for Matching University Curricula to Professional Competencies

UniSkill: A Dataset for Matching University Curricula to Professional Competencies

Summary

Research on skill extraction has mostly looked at job ads, employers and candidates. The education side, meaning which competencies university courses actually teach, has received far less attention, and no open dataset linked course learning goals to standardized job skills. UniSkill fills this gap. The authors scraped graduate-level course descriptions (2019–2025) from five Finnish universities, covering IT, Business and Generic courses. They paired these with ESCO skills belonging to two occupation groups and annotated each pair manually at two levels: the course title against a skill, and an individual course sentence against a skill. They also released annotation guidelines and an LLM-generated synthetic training set with prompting guidance. For baselines, they ranked candidates by sentence-embedding similarity and added a fine-tuned BERT classifier that checks whether each course–skill pair truly matches. Combining course sentences with titles reached 87% F1 and 89% recall, and synthetic augmentation improved performance. The resource supports course-to-skill and skill-to-course recommendation and curriculum–labour-market alignment studies.

Key points

  • The manually annotated data contains 2,192 annotations: 1,096 sentence–skill pairs and 1,096 title–skill pairs.
  • The skills are drawn from two ESCO occupation groups, Systems Analysts (2511) and Management and Organization Analysts (2421). Course text is taken from the "Content" and "Learning Objectives" sections.
  • The best BERT model scores 87% F1 and 89% recall. The evaluation set has 301 pairs (108 positive, 193 negative).
  • Training used AutoModelForSequenceClassification with binary labels, AdamW, a learning rate of 2×10⁻⁵, 7 epochs and a maximum sequence length of 128. The paper also analyses false negatives.
  • The dataset and model are released on Hugging Face as nurlanm/UniSkill_dataset and nurlanm/UniSkill_Bert.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/2n39qzvk2eqe

@inproceedings{musazade2026uniskill,
  title = {{UniSkill: A Dataset for Matching University Curricula to Professional Competencies}},
  author = {Musazade, Nurlan and Mezei, József and Zhang, Mike},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2n39qzvk2eqe},
  url = {https://doi.org/10.63317/2n39qzvk2eqe}
}

References

36 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Palakorn Achananuparp, Ee-Peng Lim, and Yao Lu. 2025. A Multi-Stage Framework with Taxonomy-Guided Reasoning for Occupation Classification Using Large Language Models. Version Number: 2. find
  2. Viktor Andonovikj, Biljana Mileva Boshkoska, Tjaša Redek, and Pavle Boškoski. 2024. A data-driven approach to aligning academic offerings with industry needs for business and economy in Slovenia. Journal of Decision Systems, 33(sup1):284– 296. find
  3. Maiia Y. Bocharova, Eugene V. Malakhov, and Vitaliy I. Mezhuyev. 2023. VacancySBERT: the approach for representation of titles and skills for semantic similarity search in the recruitment domain. AAIT, 6(1):52–59. find
  4. Jacob Cohen. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement, 20(1):37–46. find
  5. Jens-Joris Decorte, Matthias De Lange, and Jeroen Van Hautte. 2025. Multilingual jobbert for cross-lingual job title matching. find
  6. Jens-Joris Decorte, Severine Verlinden, Jeroen Van Hautte, Johannes Deleu, Chris Develder, and Thomas Demeester. 2023. Extreme Multi-Label Skill Extraction Training using Large Language Models. ArXiv:2307.10778 [cs]. — [arXiv:2307.10778](https://arxiv.org/abs/2307.10778) find
  7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… find
  8. Laura Felby, Jacob Nielsen, Anna Roesen, Morten Mathiasen, and Sune Müller. 2024. Evaluating labour market–educational programme fit: A case study on aligning supply and demand for competences using a generic approach. NJVET, 14(4):1–23. find
  9. Jibril Frej, Anna Dai, Syrielle Montariol, Antoine Bosselut, and Tanja Käser. 2024. Course Recommender Systems Need to Consider the Job Market. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 522–532. ArXiv:2404.10876 [cs]. — [arXiv:2404.10876](https://arxiv.org/abs/2404.10876) find
  10. Ann-Sophie Gnehm, Eva Bühlmann, and Simon Clematide. 2022. Evaluation of transfer learning and domain adaptation for analyzing German-speaking job advertisements. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 3892–3901, Marseille, France. European Language Reso… find
  11. Alireza Javadian Sabet, Sarah H. Bana, Renzhe Yu, and Morgan R. Frank. 2024. Course-Skill Atlas: A national longitudinal dataset of skills taught in U.S. higher education curricula. Sci Data, 11(1):1086. find
  12. Hamit Kavas, Marc Serra-Vidal, and Leo Wanner. 2025. Multilingual skill extraction for job vacancy– job seeker matching in knowledge graphs. In Proceedings of the Workshop on Generative AI and Knowledge Graphs (GenAIK), pages 146– 155, Abu Dhabi, UAE. International Committee on Computational Linguis… find
  13. Imane Khaouja, Ismail Kassou, and Mounir Ghogho. 2021. A Survey on Skill Identification From Online Job Ads. IEEE Access, 9:118134– 118153. find
  14. Seongwoon Kim, Yong-Yeol Ahn, and Jaehyuk Park. 2024. Labor Space: A Unifying Representation of the Labor Market via Large Language Models. In Proceedings of the ACM Web Conference 2024, pages 2441–2451, Singapore Singapore. ACM. find
  15. Kirsty Kitto, Nikhil Sarathy, Aleksandr Gromov, Ming Liu, Katarzyna Musial, and Simon Buckingham Shum. 2020. Towards skills-based curriculum analytics: Can we automate the recognition of prior learning? In Proceedings of the tenth international conference on learning analytics & knowledge, pages 171… find
  16. Dor Lavi, Volodymyr Medentsiy, and David Graus. 2021. conSultantBERT: Fine-tuned Siamese Sentence-BERT for Matching Jobs and Job Seekers. ArXiv:2109.06501 [cs]. — [arXiv:2109.06501](https://arxiv.org/abs/2109.06501) find
  17. Martin le Vrang, Agis Papantoniou, Erika Pauwels, Pieter Fannes, Dominique Vandensteen, and Johan De Smedt. 2014. Esco: Boosting job matching in europe with semantic interoperability. Computer, 47(10):57–64. find
  18. Hyeongjae Lee and Inho Hong. 2025. Quantifying the influence of vocational education and training with text embedding and similarity-based networks. PLoS One, 20(8):e0329405. find
  19. Florin Leon, Marius Gavrilescu, Sabina-Adriana Floria, and Alina-Adriana Minea. 2024. Hierarchical Classification of Transversal Skills in Job Ads Based on Sentence Embeddings. Information, 15(3):151. ArXiv:2401.05073 [cs]. — [arXiv:2401.05073](https://arxiv.org/abs/2401.05073) find
  20. Lightcast. 2025. Skillabi. https://lightcast.io/solutions/education/skillabi. Accessed: 2025-10-13. ↗
  21. Mary L. McHugh. 2012. Interrater reliability: the kappa statistic. Biochem Med (Zagreb), 22(3):276–282. find
  22. Nurlan Musazade, Mike Zhang, and József Mezei. 2025. Cross-lingual sentence-level skill identification in English and Danish job advertisements. In Proceedings of the 8th International Conference on Natural Language and Speech Processing (ICNLSP-2025), pages 410–415, Southern Denmark University, Ode… find
  23. Xiang Qian Ong and Kwan Hui Lim. 2023. SkillRec: A Data-Driven Approach to Job Skill Recommendation for Career Insights. In 2023 15th International Conference on Computer and Automation Engineering (ICCAE), pages 40–44, Sydney, Australia. IEEE. find
  24. Álvaro Pérez Amado. 2025. Skill-sim: a sentence-transformers model for skill similarity and job matching. https://huggingface.co/alvperez/skill-sim-model. ↗
  25. V. S. Ramazanova, A. S. Yerimbetova, M. A. Sam-betbaeva, Zh. B. Sadirmekova, and S. K. Serik-bayeva. 2024. Enhancing the Quality of Clustering Job Skills Through the Fine-Tuning of the Sentence Transformer Model. In 2024 9th International Conference on Computer Science and Engineering (UBMK), pages… find
  26. Vidya Ramnarine, Nicholas Chamansingh, and Patrick Hosein. 2025. Using Natural Language Processing to Correlate University Curricula with Required Job Skills. In 2025 International Conference on Artificial Intelligence, Computer, Data Sciences and Applications (ACDSA), pages 1–6, Antalya, Turkiye. I… find
  27. Tim Schlippe and Koen Bothmer. 2023. Skill Scanner: An AI-Based Recommendation System for Employers, Job Seekers and Educational Institutions. Int. J. Adv. Corp. Learn., 16(1):55–64. find
  28. Elena Senger, Mike Zhang, Rob van der Goot, and Barbara Plank. 2024. Deep learning-based computational job market analysis: A survey on skill extraction and classification from job postings. In Proceedings of the First Workshop on Natural Language Processing for Human Resources (NLP4HR 2024), pages… find
  29. Faisal Rashid Sheikh, Pascal Hürten, Insa Reichow, and Lutz Goertz. 2024. Learning Resources Metadata: Opportunities and Challenges in the German "Innovationswettbewerb INVITE" Program. In Proceedings of the International Conference on Dublin Core and Metadata Applications. Dublin Core Metadata Init… find
  30. Johan De Smedt, Martin le Vrang, and Agis Papantoniou. 2015. ESCO: towards a semantic web for the european labor market. In Proceedings of the Workshop on Linked Data on the Web, LDOW 2015, co-located with the 24th International World Wide Web Conference (WWW 2015), Florence, Italy, May 19th, 2015,… find
  31. Natalia Vanetik and Genady Kogan. 2023. Job Vacancy Ranking with Sentence Embeddings, Keywords, and Named Entities. Information, 14(8):468. find
  32. Zhen Xu, Xinjin Li, Yingqi Huan, Veronica Minaya, and Renzhe Yu. 2025. From Course to Skill: Evaluating LLM Performance in Curricular Analytics. ArXiv:2505.02324 [cs]. — [arXiv:2505.02324](https://arxiv.org/abs/2505.02324) find
  33. Renzhe Yu, Subhro Das, Sairam Gurajada, Kush Varshney, Hari Raghavan, and Carlos Lastra-Anadon. 2021. A research framework for understanding education-occupation alignment with NLP techniques. In Proceedings of the 1st Workshop on NLP for Positive Impact, pages 100–106, Online. Association for Compu… find
  34. Mike Zhang. 2024. Computational job market analysis with natural language processing. arXiv preprint arXiv:2404.18977. — [arXiv:2404.18977](https://arxiv.org/abs/2404.18977) find
  35. Mike Zhang, Rob van der Goot, and Barbara Plank. 2023. ESCOXLM-R: Multilingual taxonomy-driven pre-training for the job market domain. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11871–11890, Toronto, Canada. Association f… find
  36. Mike Zhang, Rob van der Goot, and Barbara Plank. 2024. Entity linking in the job market domain. In Findings of the Association for Computational Linguistics: EACL 2024, pages 410–419, St. Julian’s, Malta. Association for Computational Linguistics. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/musazade-2026-uniskill-dataset-matching-university-curricula/