A Study on the Classification of Chinese Medicine Records Using BERT, Chest Impediment as an Example.

Traditional Chinese Medicine (TCM) is the treasure of Chinese civilization and plays an indispensable role in China’s medical system, but the diagnosis of TCM relies heavily on doctors’ experience, which can affect the accuracy of diagnosis in practice. With the development of natural language processing technology, its mechanism can learn from a large amount of unstructured text to obtain a comprehensive and unified classification model. In this paper, we take chest impediment disease ( i.e. coronary heart disease in Western medicine) as an example and build a pre-training diagnostic model based on the BERT model for TCM texts to accomplish the text classification task for different types of chest impediment medical records. Its overall F1 value reached 0.851, which improved 0.096 compared with the model without TCM pre-training; it also explored the problem of long text truncation and stopwords removing of TCM cases, which improved 0.087 compared with no TCM stopwords removing. This paper introduces natural language processing into the TCM auxiliary diagnosis problem, in order to improve the informationization, standardization and intelligence of TCM in the new era.