TEXTUAL TRANSFORMATION IN KEY AUDIT MATTERS RESEARCH: A DEEP LEARNING-BASED GLOBAL BIBLIOMETRIC STUDY

Abstract

Purpose: This study provides a comprehensive bibliometric analysis of the global research landscape concerning Key Audit Matters (KAMs), with a specific focus on the "textual transformation" of audit reporting. As the implementation of International Standard on Auditing (ISA) 701 has moved from early adoption to a mature phase, research has shifted from simple quantitative counting to sophisticated qualitative assessments of audit narratives. Methodology: Leveraging a dataset of global publications (2016–2026) from the Scopus database, this study employs bibliometric mapping to visualize the field’s intellectual structure. The analysis uniquely highlights the integration of deep learning and Natural Language Processing (NLP) models, such as FinBERT and BERT, which researchers are increasingly using to decode the semantic content, tone, and readability of audit disclosures. Findings: The bibliometric mapping reveals three primary research clusters: (1) Determinants, focusing on how audit partner gender, audit committee expertise, and firm characteristics influence KAM specificity; (2) Consequences, examining the impact of KAMs on audit quality, fees, and stock price crash risk; and (3) Textual Attributes, where deep learning tools analyze "boilerplate" versus "firm-specific" language to predict financial restatements and credit risk. The results demonstrate a significant geographic expansion from developed markets like the UK and Australia to emerging economies such as China, Thailand, Jordan, and Indonesia. Originality/Value: This study is among the first to categorize the evolution of KAM literature through the lens of textual transformation. By highlighting the shift toward AI-driven analysis, it provides a roadmap for future research, suggesting that the "black box" of the audit process is increasingly being opened through the deep learning of narrative audit reports.