Research On Chinese Named Entity Recognition Based On Deep Learning

Named entity recognition refers to identifying and labeling named entities that exist in text,it has a wide range of applications,it is an important technical foundation for many tasks such as information extraction,machine translation,and knowledge graph.How to use deep learning to identify named entities from Chinese texts automatically,accurately and quickly has great research value and practical significance.The paper first describes in detail the principles and limitations of the Bi LSTM-CRF framework.At present,the Bi LSTM-CRF model based on character vectors has become popular methods,but it has limitations in terms of text representation,feature extraction,and model training efficiency.Then the paper improves on the problems existing in the current method,uses word vectors union character vectors to enrich the text representation,and proposes two improved deep learning models.One is the CNN-Bi GRU-CRF model,CNN network is added to extract the semantic spatial features of the text,and the GRU network is used instead of the LSTM network to simplify the complex structure of the model.The other is the Transformer-Bi GRU-CRF model,which extracts text features through the structure of the Transformer encoder,the Self-Attention mechanism learns the semantic information inside the text,the multi-head attention enhances the feature representation of the text,which helps to solve the problem of the text long distance dependence and keyword highlighting.Finally,the paper conducts experimental research on the proposed improved method.Experimental results show that compared with the benchmark model Bi LSTM-CRF,the recognition effect based on the CNN-Bi GRU-CRF model has increased by 1.91% in F1 value,and the GRU network performs better in terms of training speed;the improved model based on Transformer-Bi GRU-CRF shows the F1 value increased by 3.41%,showing the powerful learning ability of the Transformer model for text features.The feasibility and effectiveness of the work in the paper are verified by experiments.