Design And Implementation Of Text Processing System Based On Deep Learning

With the development of artificial intelligence technology,the legal,medical and security industries have been profoundly affected.In these industries,most of the data can exist in the form of text.The purpose of text processing is to better manage the text and obtain the information needed by users from the text,specifically,to classify the target text and extract information.Deep learning has achieved good application effects in speech recognition,computer vision and machine translation,as well as text classification and other text processing tasks.Text classification is a core part of text processing.The main task is to learn the content and labels of a given text,generate the mapping relationship into a classifier,and use the classifier to classify the text of unknown categories.The main research work of this thesis is as follows:1.In order to replace the complex feature engineering in traditional methods,the deep learning technology is applied to the text classification task,and the neural network model is used to learn the feature mapping in the text and realize the automatic extraction of text features.This thesis introduces the principle of text classification algorithm based on deep learning.The deep learning model mainly adopts convolutional neural network and hierarchical attention network.2.On the basis of the research of the two deep learning models,multiple deep learning models were fused to improve the accuracy of text classification,and the open Chinese text classification data set was used for comparative experiments.According to the analysis of the experimental results,the convolutional neural network model has the shortest training time and the lowest classification accuracy.The hierarchical attention network can improve the accuracy by 3% on this basis.However,after the fusion of the two models,the accuracy is 6% higher than that of the convolutional neural network.3.Designed and implemented a legal text processing system based on deep learning for the text processing research of the legal industry.The implementation of the system is mainly based on the deep learning framework TensorFlow,data set from the network on the acquisition of the contract template,using convolution integration model of neural network and hierarchical attention network constructing classifier,in the treatment of the contract documents as well as input from the user demand for information extraction,and contract template matching to the needs of user.On the basis of deep learning and natural language processing technology,this thesis mainly studies text processing in the legal industry.Documents in the legal industry,such as adjudication documents,contract texts,laws and regulations,are generally large in number and complex in content,with inefficient manual processing methods and low information utilization rate.In view of this situation,more efficient and accurate methods should be adopted to complete text processing tasks.