Understanding Citizen Feedback of Jakarta Government Super App: Leveraging Deep Learning Models

Jakarta Kini (JAKI) is a mobile application created by Jakarta Government to facilitate the Jakarta residents towards public services. Sentiment analysis of users' reviews should be done to provide an understanding of the essence of the issues which JAKI's users face. In this paper, word cloud analysis and a comparison of several deep learning methods were done to do the sentiment analysis of JAKI's reviews. Word cloud analysis indicates the satisfaction of the users as well as the need for several improvements that should be made by the JAKI developers. Furthermore, several deep learning techniques were employed such as LSTM, BiLSTM, GRU, BiGRU, and IndoBERT since their performance is better than conventional machine learning. Results show that the IndoBERT model outperforms another model. This shows that IndoBERT can effectively be used for this Indonesian sentiment analysis task and can be used as a reference method for analyzing Indonesian reviews in JAKI and other mobile apps in Google Play Store and Apple App Store.