Lyrics-Based Mood Detection in Music Using Text Mining Techniques

This paper classifies songs based on lyrical content for the growing demand of better user experiences and personal music recommendations. The main idea is to come up with a classifier that could determine whether a song is happy, sad, calm, or energetic based on the lyrics. It uses indepth text analysis over data retrieved from kaggle(Spotify) to see word distributions which differ across categories of emotions. Achieving meaningful text representation, feature selection uses the Bag of Words (BOW) model, along with Part-of-Speech (POS) tagging and stemming through WordNet. The classification model built with Random Forest and XGBoost uses hyperparameter tuning through grid search to improve model performance. The outcome of this research demonstrates the effective application of text mining techniques in Python to analyze, categorize, and predict the emotion of music, thus offering a more personalized music experience for users.