Emotion Classification of Song Lyrics Using the Naïve Bayes SVM Method
Abstract
In the digital era, emotion recognition in music enhances personalized recommendation systems. This study classifies emotions in Indonesian song lyrics—happy and sad—using Naïve Bayes and Support Vector Machines (SVM). A dataset of 1,051 lyrics was scraped from Genius and labeled using the NRC Word-Emotion Lexicon. Preprocessing included case folding, normalization, tokenization, stopword removal, and stemming, followed by TF-IDF feature extraction. Model evaluation used 5-Fold Cross Validation with accuracy, precision, recall, and f1-score metrics. Results show that Naïve Bayes outperforms SVM, achieving 67% accuracy and a 66% f1-score, while SVM reached only 50% accuracy and a 33% f1-score. Thus, Naïve Bayes is more effective for emotion classification in Indonesian lyrics
