本研究運用BERT模型對台灣上市公司法說會的文本進行深度情緒分析，並根據不同法規期間，將資料分為三個時期研究，旨在探討法說會文本情緒對公司股價的影響以及法說會召開頻率是否能夠減少資訊不對稱的問題。

研究結果顯示，隨著監管要求提高法說會頻率，市場對法說會情緒的反應變得越來越迅速和明確。在提高法說會召開頻率之後，法說會隔天股價出現0.18%的顯著異常報酬，這表明法說會傳遞了額外資訊，且法說會後第二天不再有顯著異常報酬。相較於無強制規範法說會召開頻率時，法說會後第十天才出現顯著-0.22%的異常報酬，資訊不對稱問題得到改善。此外，短期內負面報酬比例顯著下降，市場對法說會情緒的解讀從悲觀轉向樂觀。

本研究不僅證實了法說會情緒對股價有實質影響，也顯示了利用自然語言處理技術對法說會進行情緒分析在投資策略中的實用性和有效性。

This study employs the BERT model to conduct a deep sentiment analysis of earnings call transcripts from publicly traded companies in Taiwan. The data is segmented into three periods based on different regulatory phases, aiming to explore the impact of sentiment in earnings call transcripts on stock prices and whether the frequency of earnings calls can mitigate the issue of information asymmetry.

The results indicate that as regulatory requirements increased the frequency of earnings calls, the market's response to the sentiment expressed in these calls became more rapid and clear. Following the increased frequency of earnings calls, there was a significant abnormal return of 0.18% on the day after the earnings call, indicating that the earnings calls conveyed additional information. Furthermore, no significant abnormal return was observed on the second day after the earnings call. In contrast, during the period without mandatory frequency regulations, a significant abnormal return of -0.22% was only observed on the tenth day after the earnings call, indicating an improvement in information asymmetry. Additionally, the proportion of short-term negative returns significantly decreased, and the market's interpretation of earnings call sentiment shifted from pessimistic to optimistic.

This study not only confirms that the sentiment of earnings calls has a substantial impact on stock prices but also demonstrates the practicality and effectiveness of using natural language processing techniques for sentiment analysis of earnings calls in investment strategies.

1. Al-Ali, A. G., Phaal, R., & Sull, D. (2020). Deep learning framework for measuring the digital strategy of companies from earnings calls. arXiv preprint arXiv:2010.12418.
2. Albrizio, S., Dizioli, A., & Simon, P. V. (2023). Mining the Gap: Extracting Firms’ Inflation Expectations From Earnings Calls. International Monetary Fund.
3. Atiase, R. K., & Bamber, L. S. (1994). Trading volume reactions to annual accounting earnings announcements: The incremental role of predisclosure information asymmetry. Journal of accounting and economics, 17(3), 309-329.
4. Azimi, M., & Agrawal, A. (2021). Is positive sentiment in corporate annual reports informative? Evidence from deep learning. The Review of Asset Pricing Studies, 11(4), 762-805.
5. Ball, R., & Brown, P. (2013). An empirical evaluation of accounting income numbers. In Financial Accounting and Equity Markets (pp. 27-46). Routledge.
6. Blau, B. M., DeLisle, J. R., & Price, S. M. (2015). Do sophisticated investors interpret earnings conference call tone differently than investors at large? Evidence from short sales. Journal of Corporate Finance, 31, 203-219.
7. Brockman, P., Li, X., & Price, S. M. (2015). Differences in conference call tones: Managers vs. analysts. Financial Analysts Journal, 71(4), 24-42.
8. Brown, S., Hillegeist, S. A., & Lo, K. (2004). Conference calls and information asymmetry. Journal of Accounting and Economics, 37(3), 343-366.
9. Chapman, K. (2018). Earnings notifications, investor attention, and the earnings announcement premium. Journal of Accounting and Economics, 66(1), 222-243.
10. Chin, A., & Fan, Y. (2023). Leveraging Text Mining to Extract Insights from Earnings Call Transcripts. Journal of Investment Management, 21(1), 81-102.
11. Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805.
12. Elliott, W. B., Loftus, S., & Winn, A. (2024). To read or to listen? Does disclosure delivery mode impact investors' reactions to managers' tone language?. Contemporary Accounting Research, 41(1), 7-38.
13. Frankel, R. M., Jennings, J. N., & Lee, J. A. (2017). Using natural language processing to assess text usefulness to readers: The case of conference calls and earnings prediction. Available at SSRN 3095754.
14. Frankel, R., Johnson, M., & Skinner, D. J. (1999). An empirical examination of conference calls as a voluntary disclosure medium. Journal of Accounting Research, 37(1), 133-150.
15. Giannini, R., Irvine, P., & Shu, T. (2019). The convergence and divergence of investors' opinions around earnings news: Evidence from a social network. Journal of Financial Markets, 42, 94-120.
16. Hiew, J. Z. G., Huang, X., Mou, H., Li, D., Wu, Q., & Xu, Y. (2019). BERT-based financial sentiment index and LSTM-based stock return predictability. arXiv preprint arXiv:1906.09024.
17. Hsiao, Wei-Chung (2019). The Impacts of Conference Call Information Disclosure on Stock Market. Unpublished Master's Thesis. National Taiwan Ocean University.
18. Koval, R., Andrews, N., & Yan, X. (2023, July). Forecasting Earnings Surprises from Conference Call Transcripts. In Findings of the Association for Computational Linguistics: ACL 2023 (pp. 8197-8209).
19. Lin, Yi-Ying (2017). The Relationship between Information of Video Conference Call and Value-relevance. Unpublished Master's Thesis, National Chung Cheng University.
20. Liu, Kuan-Yu (2024). Research on Applying Machine Learning to Corporate Financial Statement Analysis and Abnormal Returns. Unpublished Master's Thesis, Ming Chuan University
21. Tasi, Hsing-Ju (2002). An Empirical Examination of Information content of Conference Calls and Management's Holding Incentives. Unpublished Master's Thesis, National Taipei University.
22. Solberg, L. E., & Karlsen, J. (2018). The predictive power of earnings conference calls: predicting stock price movement with earnings call transcripts (Master's thesis).
23. Zhao, L., Li, L., Zheng, X., & Zhang, J. (2021). A BERT based sentiment analysis and key entity detection approach for online financial texts. In 2021 IEEE 24th International conference on computer supported cooperative work in design (CSCWD) (pp. 1233-1238). IEEE.