Research On BiGRU-CapsNet Text Classification Model

Text classification is a very important part of text mining,has a huge role in the era of artificial intelligence.For text classification,the traditional method is based on the machine learning model of artificial extraction of features.However,with the continuous development of neural network structure,deep learning has made great breakthroughs in the fields of language and image,and its feature representation ability is far superior to that of artificial extraction methods.In order to improve the classification performance of text classification model,this paper introduces the general process of text classification based on traditional machine learning algorithms,as well as common text representation methods and text classification performance evaluation.This paper mainly uses the deep learning model to study text classification.The main contents are as follows:(1)TextCNN and Bidirectional Gated Recurrent Unit Network(BiGRU)model.These two models can better end-to-end classification based on the original text input,thus avoiding problems such as manual design features.After verification by a specific data set,the TextCNN and BiGRU models used in this paper can effectively improve the accuracy and AUC value compared with traditional models such as KNN,SVM,Logistic Regression and Naive Bayes.(2)BiGRU-CapsNet,a hybrid text classification model based on Capsule Neural Network(CapsNet)and BiGRU.CapsNet has a good classification effect in the image field,but it is rarely used in the field of text classification.This paper combines BiGRU and CapsNet,and proposes the BiGRU-CapsNet text classification model to extend CapsNet to the field of text classification.The model has the advantages of BiGRU and CapsNet,and enriches the feature extraction angle based on BiGRU.The experimental comparison shows that BiGRU-CapsNet has better classification performance than BiGRU model.