"The project utilizes a 6-stage Apache Spark MLlib Pipeline that transforms raw lyrics into genre predictions, demonstrating the potential of text-based features in music classification."
"The Mendeley Music Dataset contains around 28,000 songs spanning 8 genres, with data cleaning processes that include removing empty lyrics and normalizing text."
"Using TF-IDF, the project re-weights term frequencies to enhance the classification process, showcasing the importance of this classic NLP technique in genre prediction."
"The hypothesis tested was whether music genres could be classified purely from lyrics, and the results confirmed this with an accuracy of around 78%."
Music genre classification can be achieved using song lyrics instead of traditional audio features. A project was developed using a dataset of 28,000 songs across 8 genres, employing a 6-stage Apache Spark MLlib pipeline. The pipeline includes tokenization, stop words removal, feature mapping, and logistic regression for genre prediction. The approach yielded around 78% accuracy, demonstrating the effectiveness of lyrics in genre classification through Natural Language Processing techniques.
Read at Medium
Unable to calculate read time
Collection
[
|
...
]