Analisis Sentimen Ulasan Produk Kecantikan di Tokopedia Menggunakan IndoBERT

Abstract

The rapid growth of information and communication technology has significantly influenced consumer behaviour, especially in online shopping. In Indonesia, e-commerce has transformed how customers’  purchase beauty products, with platforms like Tokopedia playing an important role in facilitating these transactions. However, the increasing volume of customer reviews poses challenges for businesses to analyze consumer sentiment effectively. This research aims to analyze customer reviews of Skintific moisturizer products using Natural Language Processing (NLP) with the IndoBERT model, which is specifically trained for the Indonesian language. Using a dataset of 1,394 reviews, the research adopts a mixed-method approach, utilizing qualitative sentiment analysis for customer reviews and quantitative analysis for star ratings. The sentiment analysis process involves web scraping (ParseHub), text preprocessing, and sentiment classification with the findings show that 72,7% of reviews were positive, 17,4% negative, and 9,9% neutral. Additionally, star ratings strongly correlated with sentiment labels, where 5-star reviews were predominantly positive (85,86%) and 1-star reviews mostly negative (85,96%). The study concludes that IndoBERT effectively classifies sentiments in Indonesian-language reviews, providing valuable insights for businesses to enhance product strategies and customer satisfaction.