Echoes of Opinion: Leveraging Advanced Deep Learning for Spotify Review Classification

Abstract

The escalation of user-generated content on various streaming platforms, such as Spotify, has created invaluable insights into users’ feedback to improve user engagement and interaction. The proposed study aims to classify the Spotify user reviews and subsequently perform sentiment analysis by gathering a dataset from Kaggle containing the users’ reviews and appropriate labels. Initially, feature extraction techniques such as Bag of Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), and Part of Speech Tagging (POS) have been applied. Later, several Machine Learning, Deep Learning, and Transformer-based models have been deployed to classify the reviews. The proposed approach achieved 90 % accuracy with Support Vector Machine (SVM), while among deep learning models, Long Short Term Memory (LSTM) outperformed Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) by obtaining 95 % accuracy score. Bert large, on the other hand, surpassed Bert-uncased with 94 % accuracy. These valuable findings of the proposed approach enhance the understanding of feedback analysis for streaming services and further recommend future direction to improve Spotify’s recommendation system.