Volunteering plays a vital role in addressing societal, environmental, and humanitarian challenges. However, organizations often face difficulties in efficiently identifying individuals with the right skills for specific tasks, while volunteers struggle to find opportunities aligned with their interests and competencies. This gap underscores the need for advanced, skill-based matching solutions. This thesis presents a matching system designed to connect volunteers with suitable tasks using the European Skills, Competences, Qualifications and Occupations (ESCO) taxonomy as a standardized skill reference. The system extracts ESCO-aligned skills from unstructured text provided by both volunteers and organizations, and computes similarity scores to generate personalized recommendations. To enhance skill extraction, this thesis systematically evaluates and combines multiple techniques, including cosine similarity for matching skills, text segmentation for breaking down unstructured text into meaningful components, Named Entity Recognition (NER) for identifying relevant entities corresponding to skills or qualifications, and large language models (LLMs) such as FLAN-T5, which are tested for their ability to refine skill extraction through advanced language understanding. The resulting hybrid approach uses cosine similarity as the primary method for matching skills, with a fine-tuned FLAN-T5 model serving as a fallback to identify additional relevant skills. For task-volunteer matching, several methods are assessed, including binary vector cosine similarity, TF-IDF, fuzzy matching, and embedding-based techniques. The final system employs an overlap-based strategy that also integrates broader ESCO skills to improve both accuracy and coverage. This work contributes a flexible, scalable, and domain-aware framework that combines open skill taxonomies with modern NLP and matching techniques, advancing automated volunteer-to-task alignment and supporting more effective volunteer management.
CIVolunteer - Matching of Personal Profiles/Skills with Task Descriptions Using LLMs
Brighsh, A. (Author). 2025
Student thesis: Master's Thesis