Structured Event Representation and Stock Return Predictability

Gang Li, Dandan Qiao, Mingxuan Zheng

We find that event features extracted by large language models (LLMs) are effective for text-based stock return prediction. Using a pre-trained LLM to extract event features from n…