Abstract
Enterprise annual reports are an important carrier for the disclosure of financial status and risk information of listed companies. Compared with financial data, the disclosure content of risk information in annual reports is more and more abundant. At present, the research on enterprise risk information extraction is basically the traditional method. In order to further improve the accuracy of risk information extraction, this paper proposes a risk information extraction based on the pre-training model BERT combined with BILSTM-CRF (Bidirectional Long Short-Term Memory Network-Conditional Random Field). The model is used to experiment with the risk text in the self-built 2016 annual report of the environmental protection industry. The results show that the F1 value of this model reaches 90.26%, which is better than other traditional models.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call