Music Genre Classification Based on Song Titles with Long Short-Term Memory

This study used deep learning models to classify songs into eight genres based on their titles. The task was challenging because many titles only have three or four words, which may not clearly indicate the genre. The study used a dataset of 50,000 songs and their attributes from Kaggle, and the titles were pre-processed to be compatible with the models. The study compared different deep learning models and found that the Long Short-Term Memory (LSTM) with an extra layer and a dropout was the most accurate, with an accuracy of 42.21%. To the best of the authors' knowledge, there have been no previous studies on music genre classification using song titles, making this study unique in its field.