Unveiling the dynamics of AI applications: A review of reviews using scientometrics and BERTopic modeling

Abstract

In a world that has rapidly transformed through the advent of artificial intelligence (AI), our systematic review, guided by the PRISMA protocol, investigates a decade of AI research, revealing insights into its evolution and impact. Our study, examining 3,767 articles, has drawn considerable attention, as evidenced by an impressive 63,577 citations, underscoring the scholarly community's profound engagement. Our study reveals a collaborative landscape with 18,189 contributing authors, reflecting a robust network of researchers advancing AI and machine learning applications. Review categories focus on systematic reviews and bibliometric analyses, indicating an increasing emphasis on comprehensive literature synthesis and quantitative analysis. The findings also suggest an opportunity to explore emerging methodologies such as topic modeling and meta-analysis. We dissect the state of the art presented in these reviews, finding themes throughout the broad scholarly discourse through thematic clustering and BERTopic modeling. Categorization of study articles across fields of research indicates dominance in Information and Computing Sciences, followed by Biomedical and Clinical Sciences. Subject categories reveal interconnected clusters across various sectors, notably in healthcare, engineering, business intelligence, and computational technologies. Semantic analysis via BERTopic revealed nineteen clusters mapped to themes such as AI in health innovations, AI for sustainable development, AI and deep learning, AI in education, and ethical considerations. Future research directions are suggested, emphasizing the need for intersectional bias mitigation, holistic health approaches, AI's role in environmental sustainability, and the ethical deployment of generative AI.