Sistema Recomendador para la Asignación de Profesores en la Universidad de Cuenca

Abstract

The assignment of teachers to courses in higher education represents a critical challenge for academic management, as it directly impacts the quality of the teaching-learning process. Despite their continued use, manual assignment processes face evident challenges, such as subjectivity, lack of standardization, and a high administrative workload. In response to this scenario, this study proposes a recommender system that combines sentiment analysis, using locally adapted transformer-based language models (RoBERTuito), with mathematical optimization techniques, aiming to align teachers' competencies with specific academic requirements. To achieve this, enriched teacher profiles were developed based on historical evaluations, automatically classified student comments, and institutionally defined competencies within the framework of the Competency Pentagon. Additionally, dynamic weights were incorporated to adjust the relevance of pedagogical and technical factors according to the particularities of each academic cycle. The results obtained from the recommender system demonstrate a high correlation between generated recommendations and manual assignments, particularly in technically oriented degree programs. Moreover, program directors who participated in a pilot test positively evaluated the system, noting that it not only significantly reduces the operational workload but also establishes itself as a strategic tool with high potential for scalability and replicability across diverse educational contexts.