Unlocking Insights with NLP on Chinese Earnings Call Transcripts
Andrew Chin, Yuyu Fan, Chang Ge, Haobo Zhang · The Journal of Investing · 2025
With the growing importance of the China A-shares market to global investors, asset managers are developing new capabilities to uncover insights into these companies. Although the Chinese language may present hurdles to some investors, resourceful analysts increasingly are turning to translation tools to understand the plethora of corporate documents. In our research, we leverage transformer-based Chinese NLP models to create a wide selection of investment signals from a dataset of required company filings resulting from earnings call meetings. In analyzing the information content of these documents, we find that the transcripts are becoming more complex and more difficult to understand. We also show that the overall sentiment expressed in the transcripts is generally positive and has been increasing over time, mainly driven by company executives. From an investment perspective, our signals can differentiate between future outperformers and underperformers over horizons ranging from one to six months. Our analysis of the impact of company size on our results suggests that our features are robust across capitalization ranges. Finally, we find that complexity and sentiment features are complementary and when used together, can further enhance stock selection strategies. Ultimately, our research can help firms leverage NLP techniques to develop proprietary investment signals and strategies to outperform in the world’s second largest stock market.