UIDE Classificator: Artificial Intelligence for the Automatic Classification of Student Feedback in Higher Education

Abstract

This study designs, implements, and evaluates UIDE Classificator, a solution based on generative models for the automatic classification of open-ended student comments in higher education. A total of 3,328 comments were processed from NPS surveys administered at the International University of Ecuador (both on-site and online modalities; academic terms 2024-2 and 2025-1). The research follows an applied qualitative approach with a non-experimental, cross-sectional design. The tool employs prompt engineering, a closed institutional thematic ontology, and a knowledge file containing synonyms and local expressions to guide multi-category classification and detect non-informative comments. Validation was conducted against a human standard (eight evaluators) using performance indicators and processing time metrics. The results show an average accuracy of 96% and thematic coverage of 94%. Operationally, the analysis time was reduced by 83% (from three human hours to 0.5 hours using artificial intelligence per 1,000 comments). The study documents recurrent successes and errors—such as confusions between semantically related categories or modality inference in the absence of context—as well as the model’s high tolerance for spelling errors and informal language. Finally, the paper discusses practical implications for improving student experience management and provides recommendations for the responsible adoption of artificial intelligence in university settings, emphasizing traceability, human review, and data governance. The main contribution lies in demonstrating the feasibility and transferability of a contextualized solution for the Latin American context, leveraging generative AI to scale the systematization of student feedback with quality and efficiency.