Research on Entity recognition of Chinese place Names based on BERT

Aiming at the problems existing in the task of entity recognition of Chinese place names, such as the semantic ambiguity of the original model in the training process and the fuzzy entity boundary of extraction. This paper proposes a BERT - based model for entity recognition of Chinese place names. In this model, the text data of Chinese place names are encoded at the word level through BERT pre-training model, and the semantic information of the words is enriched according to the Chinese context, which effectively solves the problem of unclear entity boundaries. The word vector is sent to the bidirectional long and short term memory network for feature extraction. Conditional random field is used to decode and generate label sequence to extract named entity. Our experimental results show that the model can effectively improve the accuracy of Chinese place name entity recognition, and the F1 value obtained by the original model is 5.5 percentage points higher on average.