부산대학교 도서관

Named Entity Recognition is a basic and important task in the field of natural language processing. In the field of Chinese NER, the BiLSTM-CRF model has always been favored by many researchers due to its excellent information extraction capabilities. However, with the advent of the big data era, LSTM cannot be calculated in parallel, exposing the shortcomings of low training efficiency. In this paper, we focused on model training efficiency, using Bi-directional Quasi-Recurrent Neural Networks (BiQRNN) to replace BiLSTM, the experimental results show that the model we proposed achieves good results and increases the training speed of the model by 35%.