Preprocessing Audio Datasets For Machine Learning