For a long time, investors have been tracking and studying market sentiment, attempting to identify deviations in group behavior to unearth alpha. Research reports, which convey the professional analysts’ research outcomes on the market and listed companies, inherently possess a certain degree of foresight in stock opinions, making the mining of report sentiments a potentially crucial method for investors to stay ahead. With the advent of artificial intelligence in the financial market, investors’ ability to grasp analysts’ sentiments has significantly improved. From the perspective of text analysis, we can mine and depict analysts’ attitudes, achieving timely and accurate grasp of changes in market sentiment.

Datayes has a wealth of research report data and attempts to mine sentiment changes using core extraction techniques and NLP technology, providing investors with a sharp perspective. When we receive a research report in PDF format, we structure it into HTML format, including the title, main text, and paragraphs. After converting it into structured text, we extract the core viewpoints from the report summary and score each viewpoint to obtain the sentiment data of the report.

The data starts from January 2006, with a comprehensive historical record covering 97% of China A-shares. The data for day T is updated before 01:00 on T+1 day. The data includes research report information, core viewpoints (CORE_POINT), and sentiment scores (SCORE).