Research on a Clinical Data Classification Model Based on BERT, Bi-LSTM, and Attention Mechanism

In the field of medical diagnosis, patients' clinical data encompass information such as their primary symptoms, current disease progression, past treatment experiences, and examination results, which are crucial for physicians to accurately grasp the condition and formulate treatment plans. Clinical data from patients are typically recorded in the form of long-text documents. Therefore, to better analyze clinical long-text data and enhance the accuracy and efficiency of clinical diagnosis, this paper proposes a clinical long-text classification model that integrates the BERT model, named L-A-BERT. The L-A-BERT model combines the deep bidirectional representation capabilities of BERT, the long and short-term memory characteristics of the Bidirectional Long Short-Term Memory network (LSTM), and attention mechanisms to better understand and analyze clinical long-text data. The study employed real patient data for experimental analysis, and the results demonstrated that the L-A-BERT model performs exceptionally well in clinical long-text classification tasks, with the effectiveness of each component within the model further validated through ablation experiments.