Named Entity Recognition for Ancient Chinese Based on Knowledge Embedding

With the acceleration of the digitization process of ancient literature, the automatic extraction of entity information in ancient literature can enable researchers to study ancient history and literature deeply, which is of great significance to the study of ancient literature. Currently exist insufficient characterization of Chinese character information features in ancient literature entity recognition and boundary recognition problem. According to the problem, this paper propose an ancient Named Entity Recognition (NER) model based on knowledge embedded. Through the construction of entity recognition model based on training, in which the lexicon knowledge and PinYin knowledge, and explore a variety of fusion mechanism make the input information to learn the corresponding external knowledge. The model was trained in the ancient ZuoZhuan dataset and the GWPC2023 dataset with other methods. The F1 value reached the optimum compared with the baseline model and other methods. The F1 value of ZuoZhuan and GWPC2023 datasets increased by 1.7% and 1.8%, respectively. The results verified the superiority and validity of the model.