Topic Classification of Key Audit Matters in Japanese Audit Reports by Zero-shot Text Classification

Japanese listed companies must submit audit reports to the authorities. In principle, these reports must include Key Audit Matters (KAMs), which are matters that the auditors, as professional experts, have determined to be especially important in the audit of the financial statements. This paper provides a promising approach for classifying KAMs topics using zero-shot text classification, which can save time and resources in the analytical process related to audits. The experimental results demonstrate that the method using Generative Pre-trained Transformer 4 (GPT-4) achieves the highest classification accuracy.