Named Entity Recognition for Electronic Medical Records Based on Multi-Feature Fusion and Global Context Mechanism

In recent years, character-level BERT combined with BiLSTM-CRF models has achieved notable success in Chinese EMR named entity recognition (NER). However, relying solely on character sequences presents limitations, and BiLSTM struggles to effectively capture global context. To address these issues, we propose a Chinese EMR NER algorithm based on multifeature fusion and a global context mechanism. The method integrates character, pinyin, and radical features, and enhances BiLSTM's capacity for capturing global context through a global context mechanism, followed by CRF-based sequence decoding. Experimental results demonstrate that the proposed approach outperforms existing baselines on the CCKS2017 and CCKS2018 datasets, significantly improving NER accuracy in Chinese EMRs, and highlighting its potential for practical application.