Automatic Scoring of Student Feedback for Teaching Evaluation Based on Aspect-Level Sentiment Analysis

Student feedback is crucial for evaluating the performance of teachers and the quality of teaching. Free-form text comments obtained from open-ended questions are seldom analyzed comprehensively since it is difficult to interpret and score compared to standardized rating scales. To solve this problem, the present study employed aspect-level sentiment analysis using deep learning and dictionary-based approaches to automatically calculate the emotion orientation of text-based feedback. The results showed that the model using the topic dictionary as input and the attention mechanism had the strongest prediction effect in student review sentiment classification, with a precision rate of 80%, a recall rate of 79% and an F1 value of 79%. The findings identified issues that were not otherwise apparent from analyses of purely quantitative data, providing a deeper and more constructive understanding of curriculum and teaching performance.