Research On Identification Of The Chinese Named Entity Based On Deep Learning

The Named Entity Recognition(NER)task occupies a basic and key position in the field of natural language processing.Its purpose is to identify the specific entity names existing in the text.The identification results of named entities will affect the subsequent tasks of a series of natural language processing,such as information extraction,emotion analysis and knowledge map.At present,the research progress of deep learning is deepening,so it will be very important to use deep learning to solve the task of naming entity recognition.In this paper,deep learning method is used to identify named entities:(1)To solve the problem of low computational efficiency Bi LSTM the mainstream named entity recognition network model,with the increase of sequence and the weakening of long sequence modeling ability,this paper proposes a named entity recognition model based on BERT-Deep CAN-CRF.The BERT model dynamically generates the embedded representation of the word according to the context information,which can solve the polysemy problem and extract rich underlying features.Deep CAN is a deep convolutional attention network composed of three layers of convolution neural network and multi-head attention mechanism.It can extract long sequence text features and parallel calculation,both accuracy and efficiency.The model is verified experimentally on SIGHAN2006 data set,and the F1 value is 93.37%.The model(2)In view of the problem that the Chinese named entity recognition model based on character can not make use of words information,this paper introduces word segmentation task by generating anti-network idea to assist the training of named entity recognition task.The idea of generating adversarial network is used to extract the word boundary information shared by the two tasks,and the unique information of the participle task is filtered out,in which the discriminator is the decisive factor to extract the public boundary information.Therefore,a convolution attention network is proposed as discriminator.The model is experimentally verified on the SIGHAN2006 data set and the F1 value reaches 91.82%.(3)This paper develops a prototype system of Chinese named entity recognition using Easy UI Spring Boot Mysql technology,and visualizes the result of Chinese named entity recognition.Chinese named entity identification prototype system mainly includes three modules: Web front-end page module,data storage module and named entity identification module.