Sentiment-Driven Music Recommendations: A Comparative Study of ML Algorithms and Chatbot Implementation

Music serves as an individual’s steadfast companion, akin to a friend who stands by our side through various emotional states, In like manner, a diverse range of music should be accessible to resonate with the full spectrum of human emotions. Typically, music apps curate playlists based on the popularity of songs, which may not always align with our emotional state. Therefore, our objective is to enhance the user’s music discovery process by providing tailored music recommendations based on user’s current emotion. In pursuit of this objective, our study provides a comparative analysis of various machine learning algorithms, like Decision Trees(DT), Multinomial Naive Bayes (MNB), Random Forest Classifier(RFC), Gradient Boosting Decision Trees (GBDT) and k-Nearest Neighbors (KNN).We also created a sentiment analysis based chatbot that will interact with the user to detect its current emotion. We used a comprehensive dataset comprising 278k emotion-labeled Spotify songs taken from Kaggle, that was rigorously trained using above mentioned models and their performances were compared in accurately detecting user emotions. Our research contributes to finding the most adept model for inferring user emotions, and also provides an interactive chatbot interface, ultimately culminating in a more personalised and emotionally resonant music recommendation system.