A Systematic Review of Sentiment Analysis Systems Applied to Textual Data

With the rise of social networks, vast user data is generated daily. Sentiment Analysis (SA) extracts opinions and emotions from text, supporting decision-making in domains such as marketing, finance, politics, health, and education. While recent work emphasizes models and techniques, more efficient, domain-specific approaches are needed. SA performance depends on data quality, feature selection, domain context, and classifier choice. We employ the PRISMA methodology to narrow our search scope, focusing on 4,709 papers published by reputable sources such as Springer, Elsevier, IEEE, ACM, and others from 2020 to present. Through an iterative review of titles, abstracts, and full texts, we selected 34 papers relevant to our topic, including 30 research papers and 4 survey papers, for in-depth analysis. Our survey offers a comprehensive view of developments in SA, covering domains, applications, datasets, techniques, and future directions. It highlights knowledge gaps and suggests potential for multidisciplinary use and future research.