id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
ajst-31898	Lin, Dongjie	Predicting Internal Control Deficiencies from Investor Online Interaction Texts	2025	3	.pdf	application/pdf	2252	108	34	These findings imply that follow-up questioning on high-risk areas—such as revenue recognition, receivables collection, related-party transactions, expense capitalization, and remediation progress—together with evasive wording and response behavior, can leave measurable traces in interaction texts prior to statutory disclosure. The AP gain reflects enhanced ranking performance under class imbalance, indicating that text information improves the model’s ability to identify potential deficiency cases.	cache/ajst-31898.pdf	txt/ajst-31898.txt