Horizontal and Longitudinal Comparisons Among AI Subfields: A Bibliometric Perspective

TL;DR

This study uses bibliometric methods to analyze the structural evolution and differentiation of AI subfields from 2000 to 2024, revealing increased impact, collaboration, and interdisciplinary reliance.

Contribution

It introduces a multidimensional bibliometric framework and provides the first comprehensive long-term comparison of AI subfields' structural evolution.

Findings

AI impact and dissemination have increased significantly.

Subfields show distinct structural characteristics and trajectories.

Knowledge diffusion has shifted from closed to open, interdisciplinary networks.

Abstract

Recent artificial intelligence has developed rapidly with significant interdisciplinary expansion, yet existing studies often treat it as a whole, lacking systematic long-term subfield comparisons and structural analyses, thereby limiting understanding of internal differences and evolutionary mechanisms. To address this gap, we employ bibliometric methods, using expert interviews and indicator screening to construct an analytical framework. Twelve bibliometric indicators are selected across three dimensions: Impact and Dissemination, Collaboration Characteristics, and Author Characteristics. We conduct horizontal and longitudinal analyses of five subfields (AI, CV, ML, NLP, Web\&IR) from 2000 to 2024. Using CSRankings classification and a dataset of 106,622 papers, we apply violin plots, chord diagrams, and sankey diagrams to characterize structural features and evolutionary paths. Results show that these subfields have entered high-intensity knowledge diffusion: academic impact increased, knowledge dissemination accelerated, external disciplinary reliance grown, and knowledge production shifted from closed accumulation to open, interdisciplinary, multi-actor networks. On this basis, subfields exhibit significant structural differentiation: CV leads in academic impact with a task-oriented trajectory; ML shows shrinking industry collaboration but concentrated international collaboration with a relatively dispersed structure; Web\&IR is strongly industry-driven with a stable collaboration network; AI shows continuous growth; NLP remains relatively stable. Overall, this study reveals artificial intelligence evolving from unified diffusion to structural differentiation, constructs an extensible multidimensional framework, and provides a quantitative approach for understanding complex technological field evolution.