Emotion Based Music Recommendation System Using Bidirectional Encoder Representations From Transformers BERT

Abstract

newline Music information retrieval systems and music recommender systems newlineare closely related and increasingly interesting fields of research for academia newlineand industry. There has been a gap between these two fields in recent decades newlinewith limited number of studies based on various aspects of emotion in music newlinethat can benefit both these fields. An important part of this study is the music newlineclassification task for automatically categorising music based on its emotional newlinecontent. This thesis combines knowledge from two music datasets to identify newlinemusical emotions and then tries to solve the cold-start issue in recommender newlinesystem