Joint Extraction of Symptoms and Details in Chinese Medical Records Using the Overlapping Mechanism

The description of the patient's symptoms and details in the Chinese electronic medical record is an important basis for assisting clinicians to make diagnosis and treatment decisions in the clinical decision system. Therefore, it is greatly significant to extract symptoms and details. We propose an end-to-end overlapping joint extraction method to complete the entity identification and relationship extraction of symptoms and details in the medical record. By overlapping annotation mechanism, this method uses word vector features, word segmentation features and medical dictionary features in the embedding layer of BiLSTM-CRF to implement the overlapped relationship extraction between entity identification and entity of symptom and detail in one step. The experimental results show that our method achieves 80% in F1 score on the real outpatient medical record test set, which outperforms the existed state-of-the-art methods.