부산대학교 도서관

Sentiment analysis is a critical aspect of natural language processing that can provide valuable insights into customer satisfaction and opinions. Understanding sentiment, particularly in the aviation industry, can help organizations enhance their services and improve customer experiences. To advance sentiment analysis in this field, our research analyzed Twitter data related to US airlines. We employed a data augmentation technique using a pre-trained RoBERTa model to address class imbalance and applied the fastText classifier from Facebook's AI Research (FAIR) lab to gain valuable insights. Our study aimed to comprehend and enhance customer satisfaction in the aviation industry by presenting a comprehensive approach to sentiment analysis. The results not only demonstrate the effectiveness of our methodology but also provide airlines with actionable insights to proactively improve services based on real-time customer sentiments expressed on social media. This work represents a significant stride in leveraging advanced natural language processing techniques to provide superior customer experiences. Our approach achieved an impressive F1 score of 92.05% for ternary classification and 96.23% for binary classification, surpassing state-of-the-art models. This high score highlights the robustness of our methodology, establishing it as a potent tool for extracting meaningful sentiment from social media data in the airline domain.