Development of an Optimized Convolutional Neural Network Architecture for Sentiment Analysis of Movie Reviews with a Large-scale Dataset

Abstract

Sentiment analysis plays a pivotal role in deciphering opinions and emotions expressed in movie reviews, contributing to enhanced understanding and decision-making in the film industry. In this paper, we present the development of an optimized Convolutional Neural Network (CNN) architecture for sentiment analysis of movie reviews using a large-scale benchmark dataset from the IMDB database. The research delves into meticulous experimentation with hyperparameters, including variations in convolutional layers and dimensions, to optimize model performance. Preprocessing techniques, tokenization, and feature vector creation were integral steps, enabling the transformation of textual data into numerical representations for efficient analysis. The proposed methodology showcases promising results, achieving an accuracy rate of 88.92% which favourably compares to the research published using the same benchmark dataset. Notably, the fine-tuned CNN architecture demonstrates robustness and efficacy in sentiment classification, signifying potential advancements in sentiment analysis for large datasets.