Genre Classification using Word Embeddings and Deep Learning

Music genres refer to division of music into categories on the basis of interaction between artists, market forces and culture. They help to organize music into collections by indicating similarities between compositions or musicians. Automatic classification of genre is non-trivial as it is difficult to distinguish between different genres and many times the boundaries are not clearly defined and genres are overlapping. In this paper we try to classify a list of songs present on Spotify into mainly four genres - Christian, Metal, Country, Rap using their lyrics. We apply two Word Embedding techniques namely Word2Vec and Word2Vec with TFIDF (Term Frequency-Inverse Document Frequency) on the preprocessed data in order to map words of the lyrics into vectors consisting of real numbers. On application of machine learning algorithms like Support Vector Machine, Random Forest, XGBoost (eXtreme Gradient Boosting) and Deep Neural Networks on the resultant word vectors to predict the genre, we are able to achieve a mean accuracy of 61.70% and 71.05% in Simple Word2Vec and Word2Vec with TFIDF respectively with maximum accuracy of 65.0% and 74.0% using a 3 Layer Deep Learning model in case of both the techniques.