Leveraging AI for sports fan engagement: Comparing traditional and transformer-based models

Abstract

This study builds on prior work in AI-driven sentiment analysis by investigating how transformer-based models, alongside traditional methods, can be integrated into a multi-method framework to enhance fan engagement analysis in sports. Using transformer-based natural language processing (NLP) models, this research develops a framework for real-time fan sentiment analysis on social media. A case study of a basketball team at a major university demonstrates the effectiveness of the tool in classifying and interpreting fan emotions in various engagement types. Comparative analysis of AI-based models with traditional sentiment analysis methods shows the superiority of AI models in accuracy, precision, recall, and F1-score, although traditional models may exhibit higher specificity. The study provides prescriptive recommendations from varying sensitivity and specificity needs, highlights the importance of metric selection, statistical validation, and model stability in performance evaluation, and contributes to the growing body of research on AI-driven sports analytics. The findings of this study have practical implications for strategic fan engagement initiatives. Additional validation using a Twitter dataset from the FIFA World Cup 2022 supports the models’ cross-domain applicability.