学术报告 REPORT

    学术报告

    基于深度学习的信息处理成本计算方法

    返回列表

    【报 告 人】ZHENG Rong,香港科技大学商学院,副教授

    【报告时间】2022 年 5 月 19 日(周四)14:00-15:30

    【报告地点】腾讯会议 ID:648-640-819

    【报告题目】Measuring information processing cost with language predict ability: a deep learning approach

    【摘 要】 In this study, we propose a novel text information processing cost measure and use it to investigate the incremental information content of corporate disclosures. Grounded on the language predictability theory in linguistic and cognitive science research, this new measure calculates the statistical deviation of sequential word choices from the regular language patterns learnt through large samples as the information processing difficulty of human. We implement this measure with BERT language model, which is trained on a large text dataset of firms’ 10-K annual reports. We find that lower language predictability in the MD&As section of annual report is associated with higher volatility in post-filing returns, lower accuracy and greater dispersion in analysts’ earnings forecasts, and lower firms’ future earnings performance. Additionally, we find that the immediate market reactions after the 10-K filing are more negative for firms with less predictable language in their MD&As. Overall, our results suggest that language predictability can capture the disclosure processing cost of market participants and thus an effective measure of information uncertainty and information withholding in financial market.

    【个人简介】Dr. Rong Zheng is an associate professor of information systems at the HKUST Business School. He earned his doctoral degree in information systems from the Stern School of Business at the New York University. His general research interest is about realizing business value with predictive analytics. More recently, his research examines how modern AI methods can change the information environment of financial market. He is currently an associate editor at Business & Information Systems Engineering and was the associate editors for special issues at ISR and MISQ. His work has been published in such leading outlets as: Information Systems Research, Management Science, the Accounting Review and, the Communications of the ACM.