In natural language processing technology,the long computing time and limited learning ability of chinese word seg-mentation model are the problems perplexing the academic circles.A combined SACNN+CRF model is proposed,it combines the advantages of self attention mechanism,convolution neural network and CRF to complete the task of Chinese word segmen-tation.The best parameter test results show that the best hiding number and the best iteration times of SACNN+CRF model are 100 and 200.Compared with BiSTM+CRF model,MAE,RMSE and MAPE of SACNN+CRF model it increases by 32.98％,41.89％and 36.58％,respectively.The proposed SACNN+CRF model has high efficiency and high value in the ap-plication of chinese word segmentation task.
关键词
中文分词/自注意力/卷积神经网络
Key words
chinese word segmentation/self attention/convolutional neural network
维普
万方数据