Understanding Student Sentiment in Teacher Evaluations Using Large Language Models

Abstract

This study focuses on the inherent complexity of teaching evaluation analysis and understanding at the university level. We propose an approach that merges traditional evaluation methods with advanced natural language processing (NLP) techniques, in particular, sentiment analysis and Latent Dirichlet Allocation (LDA). We seek to improve the interpretation of comments given by students, overcoming the challenges associated with vast amounts of unstructured data. The innovative use of NLP techniques in the field of teaching evaluation allows not only for positive and negative opinion categorization but also, to identify latent thematic patterns which may further provide a deeper understanding of student perceptions and biases