Enhancing Fan Base Engagement through Explainable Self-Learning Sentiment Analysis

• Published in: Moratuwa Engineering Research Conference pp. 572 - 577
• Main Authors: Lye, Mikael, Wijesinghe, Nethmi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.08.2024
• Subjects: Accuracy; Adaptation models; Analytical models; Data models; Distil-BERT; Explainable AI (XAI); Fan Engagement; Fans; Multi-Class Sentiment Analysis; Self-Supervised Learning; Sentiment analysis; Social networking (online); Sports; Testing; Tokenization
• ISSN: 2691-364X
• DOI: 10.1109/MERCon63886.2024.10688963

Individuals and brands with large fan bases face difficulty in understanding fan sentiment and its potential impact on fan engagement and performance. This is particularly pertinent within fast-paced sports such as Formula 1, where fan opinions can significantly influence driver and team morale. To address the above-mentioned problem, this study proposes the use of deep learning-based sentiment analysis techniques to enhance fan base engagement. The system would act as a tool to automate the process of marketing and public relations teams by analysing textual data provided by the fan base and providing meaningful insight/reports of the fan base's emotion of the posted content. This is achieved by a novel multi-class sentiment analysis system that utilizes a fine-tuned Distil-BERT model for classification, self-supervised techniques for self-improvement, and an explainable AI (XAI) approach for interpretability. The proposed system demonstrated strong performance during testing and evaluation, achieving an overall accuracy of 82% and an F1-score of 80%. Overall, the systems components and structure focuses on utilizing the least amount of resources while also maintaining a high prediction accuracy and speed. This ultimately results in a budget friendly and robust tool that can be integrated into bigger analytics systems.