Tested both fastText word embeddings and VADER sentiment analysis tools in a CNN + LSTM model to predict the high/low valence and arousal classification of song lyrics. This model was combined with other models (audio, song features) to try to predict music emotion.
Main repo: https://github.com/Rashy98/mood_classification_Group08
Project paper: https://drive.google.com/file/d/165FORq1yRM1stYj0jgmkNOY-VqEjHNpv/view