Text Mining of Future Dividend Policy Sentences from Annual Securities Reports

The Japanese annual securities reports provide a wide range of information, including corporate performance, business risks, ESG activities, as well as dividend policies that cover both current and future dividend payout ratios. It is widely known that future dividend increases can significantly imp...

Full description

Saved in:
Bibliographic Details
Published in2023 14th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) pp. 281 - 286
Main Authors Takano, Kaito, Okada, Tomoki, Shimizu, Yusuke, Nakagawa, Kei
Format Conference Proceeding
LanguageEnglish
Published IEEE 08.07.2023
Subjects
Annual securities reports
Companies
Data augmentation
Data models
Dividend policy prospect
Machine learning
Natural language processing
Predictive models
Text mining
Online AccessGet full text

Cover

More Information
Summary:The Japanese annual securities reports provide a wide range of information, including corporate performance, business risks, ESG activities, as well as dividend policies that cover both current and future dividend payout ratios. It is widely known that future dividend increases can significantly impact stock prices, but securities analysts are limited in their ability to analyze small and mid-sized companies due to resource constraints. This study proposes data augmentation methods that utilize a topic model and BERT model to extract relevant information on dividend policies from annual securities reports, improving the accuracy of future dividend predictions. Experimental results show that our proposed method accurately extracts information about future dividends, indicating the potential for the approach to generate new investment strategies and criteria.
DOI:10.1109/IIAI-AAI59060.2023.00063