Chinese Named Entity Recognition with Integrated Boundary Features and Location Information

Chinese named entity recognition faces challenges in word segmentation and unclear boundary demarcation of Chinese entities. This paper introduces boundary features and positional information of Chinese characters, combines them with pre-trained models, and then encodes these character vectors with boundary features and positional information using a Transformer. Finally, it uses a conditional random field model for sequence labeling to complete Chinese named entity recognition. Experiments show that this algorithm can perform boundary recognition more accurately and improve the accuracy of entity recognition.