融合字形特征的中医医案命名实体识别研究

Research on named entity recognition of TCM medical records based on glyph feature

研究了中医医案症状命名实体方法。通过汉字笔画、部首、词根编码器捕获汉字字形蕴含的语义,再通过BERT模型和BiLSTM来获取上下文的语义信息,通过CRF输出最终的命名实体结果。融合字形特征的中医医案命名实体识别方法,比以往效果较好的方法F1值提高了3.0%,有效提高了中医医案症状实体识别的准确率,可为深入挖掘中医药文本信息提供参考。

In this paper,a symptom named entity method of TCM medical records is studied.Firstly,the semantics of Chinese character glyphs are captured by Chinese character stroke,radical and root encoder,then the semantic information of context is obtained by BERT model and BiLSTM,and the final named entity results are output through CRF.The proposed method improves the F1 value by 3.0%over the previous effective method,which effectively improves the accuracy of symptom entity recognition of TCM medical records and can provide a reference for deeper mining of TCM text information.