Multi Aspect Sentiment Analysis of Mutual Funds Investment App Bibit Using BERT Method

Abstract

With the rapid development of technology, an investor no longer needs to visit investment companies to make investments. Investors can conduct all investment transactions through their smartphone screens. Bibit is one investment application that can help investors invest in mutual funds. There are many reviews given by users every day, therefore, aspect-based sentiment analysis is needed to identify the aspects and sentiments of users from each review. BERT is one popular text classification method that currently has good performance. Therefore, aspect-based sentiment analysis will be carried out in this study using the BERT method with pre-trained IndoBERT on Bibit application reviews. From the multi-aspect sentiment analysis classification results, the service aspect had the highest average accuracy score of 0.92, the user satisfaction aspect had an average accuracy score of 0.87, and the system aspect had the lowest average accuracy score of 0.75. From the sentiment analysis results, the company can improve the system and service aspects of the Bibit application to provide better service & functionality.