Abstract:

Student course evaluations have long served as a primary mechanism for assessing teaching effectiveness and informing institutional decision-making in higher education. However, while numerical rating scales provide a structured and quantifiable overview of student satisfaction, they often fail to capture the depth embedded in open-ended student comments. As institutions increasingly seek data-driven approaches to enhance teaching quality and improve the overall learning experience, there is a growing need for systematic and scalable methods capable of extracting meaningful understanding from qualitative feedback. This study addresses that need by applying sentiment analysis and emotion detection techniques to student feedback collected from course evaluations over three consecutive academic years. The primary objective is to deepen our understanding of learners’ experiences and to evaluate the effectiveness of current evaluation practices. The dataset consists of anonymized student comments gathered through institutional course evaluation systems across multiple disciplines and course levels. By focusing on three consecutive years, the study enables both cross-sectional and longitudinal examination of patterns in student sentiment, allowing for the identification of stable trends as well as shifts that may correspond to pedagogical adjustments, curricular reforms, or broader contextual changes. The study employs Python-based natural language processing (NLP) techniques. The methodological framework integrates data preprocessing procedures to ensure textual consistency and analytical accuracy. Following preprocessing, lexicon-based sentiment classification methods are used to categorize comments as either objective (fact-based) or subjective (opinion-based). The study also applies emotion detection algorithms to identify specific emotional expressions embedded within student feedback. Drawing on established emotion lexicons and machine learning classifiers, the analysis detects recurrent emotional categories such as trust, joy, fear and anger. This multi-dimensional emotional mapping allows for a richer interpretation of student perspectives than simple positive–negative polarity classification. A central component of the research involves examining the relationship between identified sentiment patterns and overall course ratings. By correlating emotional indicators with quantitative evaluation scores, the study investigates how specific emotional tones influence global assessments of teaching effectiveness. Importantly, the study also critically examines the validity and limitations of current course evaluation systems. While sentiment analysis provides empirical insights into the emotional dimensions of student feedback, it also reveals potential biases. The research, therefore, underscores the importance of combining computational methods with interpretive human judgment when utilizing sentiment analysis. The findings suggest that incorporating emotion-aware analytical tools can enhance institutional understanding of student experiences beyond numerical averages. Such findings support evidence-based pedagogical development, professional training initiatives, and curriculum design strategies. Ultimately, this research contributes to the broader discourse on improving teaching quality and strengthening the teaching–learning experience.

Keywords:

Assessment, Course evaluation, Emotion detection, Learning experience, Teaching quality, Sentiment analysis, Student feedback.