Research on Named Entity Recognition of Traditional Chinese Medicine Text Based on RoBERTa-BiLSTM-CRF

Named entity recognition in the instructions of traditional Chinese medicine is very important for the standardization and intelligent processing of traditional Chinese medicine information. Aiming at the problem that it is difficult to divide the entity boundary in the instruction text and the traditional word vector is difficult to solve the polysemy of a word, this paper proposes a model based on RoBERTa-BLLSTM-CRF to improve the accuracy and efficiency of entity recognition. RoBERTa was used to encode text to capture semantic features and context information. Subsequently, the bidirectional context is modeled by BiLSTM, and the CRF layer is combined to learn and decode the label transfer rule. Experimental results show that the RoBERTa-BiLSTM-CRF model has the best performance in the named entity recognition task, and the F1 score reaches $86.25 \%$, which is significantly higher than other models. This method integrates the advantages of different neural network components, and shows strong generalization ability and recognition performance especially when dealing with complex traditional Chinese medicine texts, providing high-quality entity recognition results for the construction of traditional Chinese medicine knowledge graph.