Mood Based Classification of Bangla Music Analyzing lyrical data using Bert Algorithm

Abstract

Music mood classification is very emergent for the musician, singer, listener.But along with audio(MP3) and Video(MP4), music lyrics are also vital components.Many researches and works have already been done for English, Hindi music mood, and mood base recommendation. There are few works done with Bangla music mood and recommendation system based on Bangla lyrics. The objective of this capstone project is to classify the Bangla Music Mood and Recommendation system based on Mood along with the genre. We are classifying 5 different Bangla music categories like Romantic, sad, happy, etc. We are using a dataset consisting of 4088 Bangla song lyrics that have different Moods and genres. Our proposed model is to do multiclass classification using Bangla BERT(Bidirectional Encoder Representation from Transformers) base. The NPL(Natural Language Processing) BERT model is open-source and it helps to achieve good accuracy, that makes it great for building models.