Disease Text Classification Model Based on Knowledge Graph and Neural Network

Accurate classification of medical disease text plays an influential role in promoting the development of medical informatization. In view of the sparseness, real-time and non-standard characteristics of disease text, this study proposes a disease text classification model that combines knowledge graph and neural network. The model extracts relevant disease knowledge from the disease text through a medical domain knowledge graph. Meanwhile, it combines the performances of convolutional neural network, recurrent neural network and attention mechanism to obtain semantic features from disease text. In addition, the introduction of attention mechanism in BiLSTM (Bi-directional long short-term memory network) can improve the efficiency of extracting effective information from disease text. Finally, we can obtain a classification model trained with a combination of semantics and knowledge. Experimental results show that the proposed model achieves more remarkable classification performance compared to other disease text classification models.