Text Mining of Online Student Reviews of Postsecondary Institutions Using Large Language Models

Choosing a college for the next four years is perceived as a high-stakes and high-risk decision for students. Oftentimes, students rely on word-of-mouth (WOM) from parents, families, peers, and guidance counselors to compare college choices. As the young generation is more adapted to searching information on the internet and online communities, eWOM draws attention to scholars to understand how students utilize that information for college decisions. This paper leverages the innovative application of Large Language Models (LLMs) for text mining, aiming to extract and analyze topic-specific sentiments expressed by online reviews. In addition, it explores the qualitative differences in college ratings by analyzing the sentiment and lexicon of student reviews. The practical implications of this study are twofold. On one hand, it assesses the accuracy of LLMs in predicting topics and sentiments in comparison to manual coding, offering a robust framework for large-scale text analysis. On the other hand, it provides insights for prospective students and their families, facilitating more informed decision-making processes of college decisions.