Structured Event Representation and Stock Return Predictability

TL;DR

This paper demonstrates that structured event features extracted by large language models significantly improve the accuracy and interpretability of stock return predictions from news articles, advancing text-based financial forecasting methods.

Contribution

It introduces a novel structured event representation (SER) model leveraging LLMs and attention mechanisms for stock return prediction, offering superior performance and interpretability.

Findings

SER-based model outperforms existing text-driven models in out-of-sample prediction

Provides highly interpretable feature structures for understanding return mechanisms

Highlights the importance of structured inputs in financial text analysis

Abstract

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 news articles, we propose a novel deep learning model based on structured event representation (SER) and attention mechanisms to predict stock returns in the cross-section. Our SER-based model provides superior performance compared with other existing text-driven models to forecast stock returns out of sample and offers highly interpretable feature structures to examine the mechanisms underlying the stock return predictability. We further provide various implications based on SER and highlight the crucial benefit of structured model inputs in stock return predictability.