Examining 1D and 2D CNN Architectures in Comparison for Sentiment Analysis in Sequential Data: A Case Study of Spotify Music Reviews

Abstract

This study examines the comparative performance of one-dimensional (1D) and two-dimensional (2D) Convolutional Neural Networks (CNNs) in processing sequential data for sentiment analysis, using Spotify music reviews as a case study. Leveraging a custom dataset from Kaggle, the study examines the effectiveness of CNN architectures in extracting meaningful patterns from text input. The study integrates PyTorch and TorchText for efficient data preprocessing and model deployment. Both architectures are evaluated based on classification accuracy, computational efficiency, and ability to handle sequential dependencies. The results highlight the strengths and limitations of each method, providing insight into their suitability for similar tasks in text-based sentiment analysis. This research provides valuable guidance for researchers and practitioners working on sequential data tasks, emphasizing the role of architectural design in achieving optimal performance.