Named Entity Recognition for Chinese Electronic Medical Records by Integrating Knowledge Graph and ClinicalBERT

Xiang XuKai Ma^*

• Xuzhou Medical University, Xuzhou, China

General purpose language models often struggle to accurately identify domain specific terminology in the medical domain, leading to suboptimal performance in named entity recognition tasks. This study proposes a method for named entity recognition in Chinese electronic medical records that combines ClinicalBERT, a language model pre-trained on clinical corpora, with structured knowledge from a medical knowledge graph. To enhance semantic understanding, entity representations derived using Translating Embeddings (TransE) are incorporated into the model. In addition, the method integrates multiple character level features, including positional labels, contextual category clues, and semantic embeddings, which help improve boundary detection in Chinese texts that lack explicit word delimiters. The input texts are first annotated using the Begin, Inside, Outside, End, Single (BIOES) tagging scheme, then encoded by ClinicalBERT and passed through a bidirectional long short term memory (BiLSTM) network followed by a conditional random field (CRF) layer for label prediction. Experimental results on public datasets show that the proposed method achieves an F1 score of 89.44 percent, outperforming existing baselines and demonstrating strong effectiveness for clinical applications.