Mapping of Educational Course Descriptions to ESCO Competences Using Large Language Models

The business sector currently faces a challenge in linking the descriptions of training offers with the actual skills acquired by participants at the time of its completion, posing a barrier for both workers when choosing job offers and companies when selecting candidates based on their skills. To address this issue, the European Union recently made available a database containing the multilingual taxonomy of European Qualifications, Competences, and Occupations (ESCO), which aims to be the fundamental reference for professional integration and mobility within Europe. This taxonomy works as a dictionary that categorizes and describes more than 3000 occupations and 13,900 competences. Consequently, the objective of this research was to develop a computational system capable of processing educational information from course descriptions and mapping it to ESCO competencies. Large Language Models (LLMs), specifically GPT-4, were used to assist in this task. In terms of implementation, the user can interact with the system based on an interface with a chat-like appearance and an API that integrates and communicates with the ESCO API and GPT-4 through Flowise, a workflow framework for LLMs. To validate the system's effectiveness, an experiment was conducted with professors coordinating various courses. They tested the platform and provided feedback on its performance in mapping course descriptions to the appropriate ESCO competences. The ultimate goal of this system is to benefit universities, students, and companies. For universities, it assists in mapping course descriptions to ESCO competencies. For students, it provides a clearer understanding of the skills they will gain from a specific course, assisting in career planning and personal development. For companies, it offers a reliable way to assess the skills of graduates, which improves hiring decisions.