A method of Named Entity Recognition in Classical Chinese based on Bert-Ancient-Chinese

Abstract

At present, in the field of Natural Language Processing (NLP), the accuracy of pre-trained models for downstream tasks such as Named Entity Recognition (NER) is significantly improved. Named Entity Recognition is a key technology in Knowledge Graphs (KG), but there are relatively few studies on this technology in the field of Classical Chinese. This paper proposes a Classical Chinese Named Entity Recognition model based on bert-ancient-chinese, which is based on bert-ancient-chinese+Recurrent Long Short-Term Memory+Conditional Random Field (BAC+RLSTM+CRF) for Named Entity Recognition and tested on the C-CLUE dataset. Compared with the baseline of the dataset, the F1 value is improved by about 22.11%.