Chinese Named Entity Recognition Method in History and Culture Field Based on BERT

Abstract With rapid development of the Internet, people have undergone tremendous changes in way they obtain information. In recent years, knowledge graph is becoming a popular tool for public to acquire knowledge. For Chinese history and culture, most researchers adopted traditional named entity recognition methods extract information from unstructured historical text data. However, method has certain defects, it easy ignore association between entities. To entities large amount cultural more accurately efficiently, this paper proposes one model combining Bidirectional Encoder Representations Transformers Long Short-Term Memory-Conditional Random Field (BERT-BiLSTM-CRF). First, BERT pre-trained language used encode single character vector representation corresponding each character. Then Memory (BiLSTM) layer applied semantically input text. Finally, label with highest probability output through Conditional (CRF) character’s category. This uses (BERT) replace static word vectors trained way. comparison, can dynamically generate semantic according context words, which improves ability vectors. The experimental results prove that proposed achieved excellent task field culture. Compared existing identification methods, precision rate, recall $$F_1$$ F 1 value been significantly improved.