Using LFtext-TextCNN to classify short text of TCM symptoms.

TCM (Traditional Chinese Medicine) symptoms are divided into six categories, and these symptom categories contain a lot of clinical diagnostic information. At present, the classification of TCM symptoms is mainly based on the experience of TCM physicians with there are many errors. The average length of the text description information of TCM symptoms is about 4, and labeling these symptom data is a very expensive task. Therefore, TCM symptom classification is a short text classification task with few samples. The current short text classification methods are difficult to effectively extract professional semantics and unique expressions in TCM symptom texts. Our method is based on LFtext-TextCNN to extract the general semantic information of TCM symptoms and the special semantic information of TCM, which are fused to linear network to classify TCM texts. Compared with the other 7 baseline models, our method is the best in TCM symptom classification task.