基于Bert-BiLSTM-CRF的中医文本命名实体识别

Named entity recognition of Chinese medical text based on Bert⁃BiLSTM⁃CRF

命名实体识别是中医智能化发展的基石。针对中医文本数据挖掘中实体识别困难的问题,构建了基于Bert-BiLSTM-CRF的命名实体识别模型。通过Bert模型在字向量的构建过程中融入注意力机制,利用BiLSTM进行特征提取,并将特征输入CRF模型之中完成最终的训练。实验选取《伤寒论》作为训练集以及测试集,划分症状、疾病名称、方剂、中药名、时间5类实体。测试集上的结果表明,该模型的准确率为96.94%,召回率为93.14%,F值为95%,命名实体识别精度较高,可以将该模型用于实际问题中。

Named entity recognition is the cornerstone of intelligent development of traditional Chinese medicine.Aiming at the problem of entity recognition in traditional Chinese medicine text data mining,a named entity recognition model based on Bert bilstm CRF was constructed.Through the Bert model,the attention mechanism is integrated into the word vector construction process,and the bilstm is used for feature extraction,and the features are input into the CRF model to complete the final training.The experiment selected treatise on febrile diseases as training set and test set,and divided into five types of entities:symptoms,disease name,prescription,Chinese medicine name and time.The results on the test set show that the accuracy rate of the model is 96.94%,the recall rate is 93.14%,and the F value is 95%.The named entity recognition accuracy is high,and this model can be used in practical problems.