Named Entity Recognition in Traditional Chinese Medicine Clinical Cases Combining BiLSTM-CRF with Knowledge Graph

Named entity recognition in Traditional Chinese Medicine (TCM) clinical cases is a fundamental and crucial task for follow-up work. In recent years, deep learning approaches have achieved remarkable results in named entity recognition and other natural language processing tasks. However, these methods cannot effectively solve the problem of low recognition rate of rare words, which is common in TCM field. In this paper, we propose TCMKG-LSTM-CRF model that utilizes knowledge graph information to strength the learning ability and recognize rare words. This model introduces knowledge attention vector model to implement attention mechanism between hidden vector of neural networks and knowledge graph candidate vectors and consider influence from previous word. The experiment results prove the effectiveness of our model.