As one of the tasks in Natural Language Processing (NLP), sentiment analysis enables computational understanding of sentiment from textual public opinion data. With the rapid growth of public opinion data available online, sentiment analysis offers a ...
As one of the tasks in Natural Language Processing (NLP), sentiment analysis enables computational understanding of sentiment from textual public opinion data. With the rapid growth of public opinion data available online, sentiment analysis offers a cost-effective and potentially real-time alternative for evaluating government services. Although research on sentiment analysis has increased significantly in Indonesia in recent years, it still faces major challenges related to linguistic diversity, limited datasets, and low representation in global NLP datasets. While the interest in this field is growing, and the strategic alignment with Indonesia’s National Artificial Intelligence Strategy 2020-2045, to our knowledge, no review study has yet examined how sentiment analysis has been applied in Indonesia’s government services domain. This study fills that gap by performing a systematic literature review of studies published between 2021 and 2025, following the PRISMA framework guidelines across Scopus, ScienceDirect, and IEEE Xplore databases. A total of 107 studies met the inclusion criteria and were reviewed deeper to synthesize methodological trends, dataset sources, and domain applications of sentiment analysis in this field. The review finds machine learning approaches such as Naïve Bayes and SVM remain dominantly used in current practice, while deep learning models like LSTM and BERT start to grow. Twitter is often used as a primary data source, and the health domain, especially COVID-19-related studies, is the largest distribution of studies. Key challenges were found, including data imbalance and representativeness limitations, limited Indonesian government-specific linguistic resources, the presence of bot-generated data, a lack of service integration, and the absence of studies that measure sentiment analysis impact in real settings. This study contributes by identifying methodological and contextual gaps that can inform researchers in future sentiment analysis research in Indonesia’s government service sector.