Aspect-Based Sentiment Analysis of Avatar 2 Movie Reviews on IMDb Using Support Vector Machine

¹ Faculty of Computer Science, Jember University
² Faculty of Computer Science, Jember University
³ Faculty of Computer Science, Jember University

^* Corresponding author: priza@unej.ac.id

Abstract

In the digital age, IMDb plays a crucial role in influencing audience movie choices. However, IMDb's movie ratings lack detailed information about specific aspects of films considered important in the industry's evaluation of audience responses. To address this gap, we conducted aspect-based sentiment analysis on 3198 reviews of Avatar 2. We focused on narrative and cinematic elements in the movie reviews, such as character, conflict, location, time, mise-en-scene, cinematography, editing, and sound. After data collection, we labeled the aspects and sentiments, and through TF-IDF weighting and SMOTE balancing, we performed sentiment classification. The Support Vector Machine model with SMOTE proved most effective, highlighting crucial features often discussed by audiences in both positive and negative sentiments. This analysis provides valuable insights for the film industry, aiding in better movie production, marketing, and a deeper understanding of audience preferences. Our research demonstrates the significance of aspect-based sentiment analysis in guiding future film-making endeavors.