Analisis Sentimen Berbasis Aspek Marketing Mix 4P pada Ulasan Aplikasi E-Wallet Dana Menggunakan Bert

Abstract

This research implements the Aspect-Based Sentiment Analysis (ABSA) method using the Bidirectional Encoder Representations from Transformers (BERT) model in analyzing user reviews related to the DANA digital wallet. DANA digital wallet is one of the popular electronic payment services in Indonesia. The purpose of this research is to understand user sentiment related to specific aspects of using DANA digital wallet and identify the accuracy ofsentiment analysis on each aspect. The ABSA method with BERT is used to recognize the aspects discussed in user reviews and classify the associated sentiments, such as product, price, place, and promotion aspects. This analysis provides insight into user preferences for the features and experiences of using the DANA digital wallet contained in their reviews. The results of this study show that aspect-based sentiment analysis conducted using BERT with the "indobenchmark/indobert-base-p2" model gets different accuracy in each aspect. The highest accuracy is obtained in the product aspect (82%), followed by price (72%), place (80%), and promotion (73%).