Pendekatan Komputasi dalam Sentiment Analysis Produk pada Media Sosial

Abstract

The rapid growth of social media has generated large amounts of user-generated content containing opinions about products and services. This condition encourages the use of sentiment analysis to extract useful insights for business decision-making. Along with advances in Artificial Intelligence and Natural Language Processing, computational approaches to sentiment analysis have evolved from lexicon-based methods, machine learning, deep learning, transformer-based models, to Large Language Models (LLMs). This study aims to analyze the development of computational approaches in product sentiment analysis on social media using a Narrative Literature Review method. The review synthesizes 25 relevant scientific articles from reputable national and international journals. The findings show that transformer-based models, especially BERT and its variants, generally provide better classification performance than traditional machine learning and deep learning methods because they can capture contextual meaning more effectively. In addition, LLMs such as GPT show strong potential for analyzing complex and limited data. This study provides an overview of the evolution, strengths, limitations, and future research opportunities of computational approaches in sentiment analysis.