BBACNet: A BERT based Cascaded Network for Emergency Department Triage

In this paper, a deep learning-based Chinese Emergency Department Triage algorithm is investigated, aiming to improve the triage efficiency and accuracy. For the complexity and real-time requirements of Emergency Department Triage, this paper proposes a proposed improved model based on BERT: BBAC (BERT-BiLSTM-attention-CNN). The model utilizes the BERT pre-training model for semantic understanding of emergency text, followed by the BiLSTM-ATT-CNN algorithm model to further extract contextual information and local key features, and finally realizes automatic triage of emergency cases. The experimental results on the emergency care dataset show that the algorithmic model has high accuracy and stability in the emergency care triage task, and can effectively assist doctors in making fast and accurate triage decisions. The contribution of this paper is to propose a new algorithmic model for emergency medical triage, which provides new ideas and methods for emergency medical informatization and intelligence. Experimental comparisons were conducted in the self-made medical text dataset, and the experimental study showed that the algorithmic model proposed in this paper achieved 87.85% accuracy, which is higher than the accuracy of the baseline model BERT.