<p>This article applies natural language processing (NLP) to extract and quantify textual information for predicting stock performance. Utilizing an extensive dataset of Chinese analyst reports and employing a customized BERT deep learning model for Chinese text, the study categorizes the sentiment of these reports as positive, neutral, or negative. The findings highlight the predictive power of this sentiment indicator for stock volatility, excess returns, and trading volume. Specifically, analyst reports with strong positive sentiment are associated with increased excess returns and intraday volatility. Conversely, reports with strong negative sentiment also heighten volatility and trading volume but lead to a decline in future excess returns. Notably, the magnitude of the effect is more pronounced for positive sentiment reports compared to negative ones. This article contributes to the empirical literature on sentiment analysis and the stock market’s response to news, particularly within the context of the Chinese stock market.</p>