2026-04-06 星期一
Speaker：Assistant Prof. William (Hongmin) Du
University of Massachusetts Boston
时间：2026年3月17日，10：00-15：00
地点：中国科学院大学玉泉路校区办公楼第一会议室
Title: Designing a Topic-Based Artifact for Corporate Disclosures: An Application to Cybersecurity Risk
Abstract:
Cybersecurity risks are disclosed in firms’ annual reports but are often embedded in lengthy and heterogeneous risk factor sections that limit interpretability. Common approaches based on keyword counts or sentiment applied to raw text are sensitive to boilerplate language and redundancy. A topic-based framework is introduced to extract and summarize cybersecurity disclosures from Item 1A of 10-K filings. The framework identifies coherent disclosure topics, filters cybersecurity-related content, and produces concise summaries that reduce redundancy while preserving topical coverage. Three summarization variants—inter-topic, intra-topic, and cybersecurity-constrained intra-topic summarization—are evaluated along four dimensions: readability, cybersecurity relevance, redundancy, and topical coverage. To assess whether summarized disclosures retain risk-relevant information, validation is conducted using an external measure of realized cybersecurity exposure derived from firms’ software stacks and associated vulnerabilities. Disclosure signals constructed from summarized text exhibit stronger associations with subsequent vulnerability exposure than baseline measures based on disclosure length or sentiment, with intra-topic summarization showing the most consistent performance. The findings demonstrate the value of targeted summarization for improving the measurement of unstructured cybersecurity risk disclosures.
个人简介：
William (Hongmin) Du is an Assistant Professor in the Management Science and Information Systems department at the University of Massachusetts Boston. He received his Ph.D. in Accounting Information Systems from Rutgers Business School and earned a Bachelor of Science in Accounting from the University of Texas at Dallas. His research focuses on cybersecurity disclosures, financial reporting, audit analytics, and the use of machine learning, large language models, and other emerging technologies in the business field.
More broadly, his work explores how new analytical tools can improve the way organizations process information and assess risk, particularly in areas such as cybersecurity reporting, auditing, and healthcare analytics. At UMass Boston, he teaches courses related to accounting analytics, information systems, cybersecurity assurance, and data analytics.
