Multi-Class Classification of Different Region Pop Songs using Spotify Database

Music Information Retrieval is a field that is interested in extracting information from musical compositions whether it is songs, instrumental pieces or scores of various genres. With the development of Machine Learning (ML) and Artificial Intelligence (AI) techniques, better models are used to infer songs and scores. These inferences are later used in detecting copyright breaches, categorizing songs as per user, developing music recommending applications and many more. Though our understanding of music has increased tremendously, linguistic based research has been limited. The work in this paper focuses on analyzing data-sets of songs of different Indian linguistic communities on the basis of some predefined musical parameters like speechiness, loudness, acousticness etc. The only similarity is the songs in comparison of different languages have the same mood or valence (positiveness). This work will help to understand how different linguistic music having the same mood, has different musical features. The method used for this project is Multi Class Neural Network for training and thus testing the data-set. The data-set is collected from a popular music app using the inbuilt Application Programming Interface (API) of the app.