Experience is all you need: a large language model application of fine-tuned GPT-3.5 and RoBERTa for aspect-based sentiment analysis of college football stadium reviews

Despite extensive research in service industries, fine-grained explorations of customer experience within spectator sports remain limited. This study pioneers a transfer learning approach grounded in a post-positivism paradigm, underpinned by an integrative customer experience framework for aspect-based sentiment analysis of user-generated content, shedding light on the complexity of the college football game day experience. Three fine-tuned large language models were employed to qualitatively identify and quantitatively analyze customer experience from Tripadvisor reviews on college football stadiums. Our findings indicated that fans' positive reactions to stimuli related to core (game dynamics), functional (facilities/services), emotional (intense feelings), and socialization (fan interactions/bonding) significantly increased the likelihood of them giving a five-star rating. Mitigating negative experiences across functional, emotional, socialization, safety, and monetary experience was crucial for achieving a top rating, with reducing negative functional issues and safety concerns having the greatest positive impact. Our study contributes to the sport management literature by establishing a unified view of customer experience, enabling a holistic conceptualization and operationalization of customer experience in spectator sports. Empirically, our research proposes targeted strategies to manage customer experience in college football and offers sport management professionals ready-to-use large language models along with detailed deployment guidelines tailored for distinct use cases.