An Extraction Model Based on RoBERTa-BiLSTM-CRF for Chinese Financial Event

Abstract

Event Extraction is one of the important research tasks of Information Extraction in Natural Language Processing. It tries to extract information from a large amount of chaotic data and presents information in a structural form. The existing Chinese event extraction methods have the inaccuracies of Chinese word segmentation, which will directly lead to incorrect identification of Chinese financial entities, affecting the accuracy of event element extraction. This paper takes Chinese financial event extraction as a sequence labeling task. It proposes an event extraction model based on PreTraining Model, Bidirectional Long-Short Term Memory Network, and Conditional Random Field. Additionally, this paper constructs the Chinese financial event dataset FinEE. At the same time, financial events are filtered from public dataset DuEE to construct dataset DuEE_Fin. As the experimental results show that the proposed Chinese financial event extraction model Roberta-BilSTM-CRF has improved accuracy, recall rate, and F1 score compared with existing models on FinEE and DuEE_Fin datasets.