A Topic Modeling－Based Analysis of the Knowledge Structure and Trends in AI－Convergence Accounting Research

Abstract

[Purpose]This study treats Korean research that addresses artificial intelligence (AI) and accounting together as a single subject domain and asks whether this domain, in which multiple disciplines participate, has in fact achieved interdisciplinary integration. To this end, it empirically analyzes the knowledge structure across the participating disciplines at three layers：topics, knowledge bases, and citation impact. [Methodology]A total of 241 articles published between 2018 and 2026 were collected from the Korea Citation Index (KCI) without restricting academic disciplines. Topics were observed through document－embedding－based topic analysis, knowledge bases through bibliographic coupling of references, and citation impact through citation counts converted into within－year percentile ranks. Robustness was examined across six samples that varied in indexing status, search－term scope, and publication period. [Findings]The topic distributions of the two largest disciplines, accounting (53 articles) and law (52 articles), were virtually non－overlapping (Cramér’s V＝.793), and only 27 of the 3,144 unique references cited by the two disciplines were shared (overlap coefficient＝.0179), so no sharing of the knowledge base was confirmed. This structure held across all six samples, and the gap widened to 13.8－fold in the narrow sample retaining only AI－related terms. No evidence of a hierarchy was found in citation impact, and temporal comparisons did not show a clear narrowing of the descriptive gap. [Implications]These findings suggest that Korean AI－accounting research is closer to a multidisciplinary juxtaposition of two comparable disciplines than to interdisciplinary integration. This study presents a design that shifts the unit of bibliographic coupling to the academic discipline in order to measure cross－disciplinary knowledge sharing directly, and proposes knowledge－base coherence as an indicator for identifying targets of interdisciplinary research support.