Medical Entity Recognition Based on BiLSTM with Knowledge Graph and Attention Mechanism

Considering the characteristics of non-standard expression in Internet text, some information extraction models have employed knowledge graphs of different areas to improve entity recognition performance. However, these existing models merely use background knowledge in a single step, which leads to insufficient use of additional clues in the knowledge graph. Aiming at the above problems, this paper proposes a new entity recognition model based on BiLSTM with knowledge graph and attention mechanism and applies the model to extract medical entities from Internet medical consultation text. The model extracts the conceptual features and candidate knowledge sets from the medical knowledge graph. The encoded concept features are embedded into the input of BiLSTM to enhance the semantic expression of words. Furthermore, the candidate knowledge sets are integrated into the attention mechanism to capture the important information in the knowledge graph and context. To verify the effectiveness of the model, we extracted medical named entities from the consultation text of Haodafu. Experimental results show that the proposed model can effectively improve the performance of entity recognition tasks.