Sentiment Analysis in Digital Spaces: An Overview of Reviews
This work addresses methodological quality issues in sentiment analysis reviews for researchers and practitioners, but it is incremental as it builds on existing review syntheses.
The authors tackled the problem of overlooked validity and scientific practices in sentiment analysis systematic reviews by synthesizing 38 reviews covering 2,275 primary studies, finding diverse applications but limited reporting rigor and challenges over time.
Sentiment analysis (SA) is commonly applied to digital textual data, revealing insight into opinions and feelings. Many systematic reviews have summarized existing work, but often overlook discussions of validity and scientific practices. Here, we present an overview of reviews, synthesizing 38 systematic reviews, containing 2,275 primary studies. We devise a bespoke quality assessment framework designed to assess the rigor and quality of systematic review methodologies and reporting standards. Our findings show diverse applications and methods, limited reporting rigor, and challenges over time. We discuss how future research and practitioners can address these issues and highlight their importance across numerous applications.