<![CDATA[Digital media platforms shape public perception of music through song lyrics and audience comments. This study analyzes emotions expressed in the lyrics and YouTube comments of Taylor Swift’s “Fortnight” using deep learning models. The dataset consists of 42 lyric lines and 13,406 user comments collected from April to December 2024. Emotion labeling was manually performed based on Plutchik’s eight basic emotions with an additional neutral category. This research applies two models: Long Short-Term Memory (LSTM) and DistilRoBERTa, with random oversampling to address class imbalance. Performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrices. As a single-song case study, this research provides a focused comparison of sequential and transformer-based architectures for simultaneous emotion analysis of lyrics and audience responses. The results show that DistilRoBERTa achieved higher accuracy (94.07%) than LSTM (90.75%), indicating the advantage of contextual transformer models in capturing nuanced emotional expressions within this dataset. However, the findings are limited to the thematic characteristics of this single-song dataset and should be interpreted within this contextual scope.]]>
Copyrights © 2026
