<![CDATA[Music is a universal medium for expressing emotions, with song lyrics serving as a narrative component rich in affective content. This study aims to analyze the emotional landscape within popular English song lyrics collected from the Spotify platform and to examine the effectiveness of Natural Language Processing (NLP) approaches in classifying these emotions. The research corpus consists of 57,494 randomly collected song lyrics without genre restrictions. Through a comprehensive analytical pipeline---ranging from text preprocessing (case folding, normalization, cleaning, tokenization, filtering, stemming), custom lexicon-based emotion labeling, TF-IDF feature extraction, to classification using a Random Forest model---the study reveals two key findings. Empirically, song lyrics are dominated by positive emotions, with romantic (36.2%) and happy (26.2%) emerging as the main themes, followed by sad (16.3%), while angry expressions (4.4%) appear least frequently, indicating significant class imbalance. Methodologically, the proposed model demonstrates solid performance with an overall accuracy of 83.03% and a weighted avg F1 score of 0.82. However, analysis of the confusion matrix and classification report uncovers performance disparities across emotion classes: angry and energetic emotions exhibit low recall (42% and 62%, respectively), likely due to imbalanced data distribution and lexicon limitations in capturing context. In conclusion, this study not only succeeds in mapping the dominance of love- and happiness-related themes in popular song lyrics but also demonstrates that classical NLP models can achieve competitive performance. The findings additionally highlight the importance of addressing class imbalance and developing more context-rich emotion lexicons or employing deep learning models in future research, in order to capture the emotional spectrum of lyrics more evenly and comprehensively.]]>
