PREDICTION OF SONG GENRE BASED ON SONG LYRICS USING MACHINE LEARNING ALGORITHMS
DOI:
https://doi.org/10.24867/21BE10JovanovicKeywords:
machine learning in music, deep learning in music, song genre predictionAbstract
The world of music is in the process of exponential growth, especially in the past ten years when a significant expansion took place thanks to the digitalization of the world. The number of new artists and songs is constantly increasing, which has led to the need to create an intelligent and efficient way for listeners to navigate the myriad of choices. One of the proven ways to filter tracks is to filter based on song genre. This paper will describe a system for classifying songs into different genres based on the song's lyrics. Data for training and testing the system is collected from several different sources and transformed into a single data set. The best results were achieved using GloVe word embeddings and LSTM network architecture. This model achieved 78% accuracy, 78% precision, 77% recall, and 77% F-measure.
References
[2] Medium https://www.apple.com/apple-music/ [datum pristupa 18.09.2022.]
[3] Alexandros T. 2017. Lyrics-based music genre classification using a hierarchical attention network
[4] LyricFind https://www.lyricfind.com/ [datum pristupa 18.09.2022.]
[5] iTunes Search API https://developer.apple.com/library/archive/documentation/AudioVideo/Conceptual/iTuneSearchAPI/index.html [datum pristupa 18.09.2022.]
[6] Ciao Luiggy R., et. al. 2019. Combining Diverse Models for Lyrics-based Music Genre Classification
[7] Michael Fell, Caroline Sporleder, Lyrics-based Analysis and Classification of Music
[8] Wu, Chuhan & Wu, Fangzhao & An, Mingxiao & Huang, Jianqiang & Huang, Yongfeng & Xie, Xing. (2019). NPA: Neural News Recommendation with Personalized Attention.
[9] Lyrics mode https://www.lyricsmode.com/ [datum pristupa 18.09.2022.]
[10] Million Song Dataset http://millionsongdataset.com [datum pristupa 19.09.2022.]
[11] Genius https://genius.com [datum pristupa 19.09.2022.]
[12] NLTK https://www.nltk.org/ [datum pristupa 19.09.2022.]
[13] GloVe: Global Vectors for Word Representation https://nlp.stanford.edu/projects/glove/ [datum pristupa 19.09.2022.]
[14] Keras https://keras.io/ [datum pristupa 19.09.2022.]
[15] Multi-class Classification: Extracting Performance Metrics From The Confusion Matrix https://towardsdatascience.com/multi-class-classification-extracting-performance-metrics-from-the-confusion-matrix-b379b427a872 [datum pristupa 19.09.2022.]
