2025/26

Contextual Representations for Matching Job Title with Skills

Aim

A job candidate spends a considerable amount of time understanding the skills required in a job advert. The project aims to develop an ML-driven search assistant that identifies the necessary skills for each job posting and guides the candidate to modify their CVs accordingly. This tool will help candidates explore the relevant skills to improve for a specific job role, while also assisting employers in crafting cleaner skill-based requirements and finding better-matched applicants.

Objectives

- To expand and represent the skills and job titles using Traditional and Neural embedding models - To embed the ESCO ontology in Neural Vector Space - To predict the relevant skills given the job title leveraging ESCO ontology

Deliverables

- Project proposal - Interim report - Dissertation - Project code-base

Research Questions

- To what extend Neural representation-based model improve the matching performance over traditional models? - To what extend ESCO ontology contributes to Job title with Skill matching problem

Methodology

- To represent the skills and job titles using Traditional and Neural embedding models - To expand the skills and job titles using Traditional and Neural approaches and ESCO ontology - To predict, test, and evaluate how Natural Language Processing (NLP) and Large language models (LLMs) can predict relevant skills for a given job title with/without context

Evaluation

This project will be evaluated using the TalentCLEF 2025 and 2026 shared task dataset on Task B: Job Title-Based Skill Prediction

Prerequisites

- Hands-on experience of Natural language processing - Hands-on experience of Data analytics and machine learning - Proven experience of Large Language Models

References

1. https://aclanthology.org/2023.findings-eacl.163/ 2. https://link.springer.com/article/10.1007/s00521-020-05302-x 3. https://talentclef.github.io/talentclef/docs/talentclef-2025/motivation/ 4. Zbib, R., Lacasa, L. A., Retyk, F., Poves, R., Aizpuru, J., Fabregat, H., … & García-Casademont, E. (2022). Learning Job Titles Similarity from Noisy Skill Labels. arXiv preprint arXiv:2207.00494

Supervisor
Ullah, Md Zia
Contact Supervisor
Required Skills
Python Machine Learning Data Analysis
Project Information
  • Academic Year 2025/26
  • Published 05 Feb 2026
  • Last Updated 05 Feb 2026