Student opinion mining: Automated topic extraction from student feedback

Czimbalmos, Olivér and Szántó, Zsolt and Kőrösi, Gábor and Becsei, Péter and Udvari, Beáta and Farkas, Richárd (2026) Student opinion mining: Automated topic extraction from student feedback. ANNALES MATHEMATICAE ET INFORMATICAE, 63. pp. 42-54. ISSN 1787-6117

Preview

Text 42_54.pdf - Published Version Download (1MB) | Preview

Abstract

While Student Evaluation of Teaching Work (SETW) is a cornerstone of quality assurance in higher education, the high volume and linguistic complexity of unstructured qualitative feedback often lead to its underutilization in institutional decision-making. This study addresses this “analysis gap” by developing an automated pipeline to process 34,000 unique Hungarian student responses from a major research university. To empower academic administration and faculty leadership with the ability to uncover latent thematic patterns within these responses, we propose a hybrid NLP framework that utilizes a Large Language Model (LLM) for thematic reclassification and Aspect-Based Sentiment Analysis (ABSA), combined with an unsupervised layer for fine-grained latent topic discovery using transformer-based embeddings and HDBSCAN clustering. Our proposed pipeline successfully identified new granular latent topics – such as lecture pacing, traceability, material accessibility and slides – providing a level of diagnostic detail that remains invisible to standard quantitative metrics. The results prove that modern natural language processing (NLP) techniques can effectively transform raw, unstructured student narratives into objective, actionable diagnostic tools.

Actions (login required)

Edit Item