Research And Application Of Text Classification Method Based On Deep Learning

Internet technology has changed the way people exchange information.Sharing consumption feelings,evaluating service quality and expressing personal opinions have become the daily habits of many users.The text information of these comments often contains the emotional tendency of users towards the current evaluation objects.Effective mining and utilization of evaluation information is an indispensable means to enhance the competitiveness of enterprises in the era of big data,and it's core technology is text classification.Text categorization has been widely used in most scenarios of natural language processing,gradually changing from a knowledge-based approach to one based on statistics and machine learning.The traditional text classification methods based on machine learning have many problems,such as complex feature engineering and high dimensional sparse text vectorization representation.In recent years,the text classification method based on deep learning has overcome the shortcomings of traditional machine learning methods to some extent and improved the accuracy of text classification.Therefore,based on the comparative analysis of the mainstream methods of text classification,this paper proposes a CN&AT-BLSTM model that integrates the attention mechanism and CNN convolutional neural network,and makes use of the advantages of long and short term memory network(LSTM)in serialization processing,and improves the accuracy of Chinese text classification by virtue of the long distance information dependence of sequences.On this basis,an experimental system is built for the application scenarios of hotel service evaluation.The specific research contents of this paper include:(1)The process of text classification generally adopt a unified morpheme assignment approach and led to the important lost or not important morpheme redundancy problem,in order to solve this problem,based on Attention mechanism and BLSTM model,designed the AT-BLSTM model,the model to focus on important morpheme in text categorization,improved the accuracy of text classification.(2)In the process of text classification,short text features are sparse and feature extraction effect is poor.To solve this problem,a CN&AT-BLSTM model is designedbased on the CNN model of convolutional neural network and the AT-BLSTM model,which effectively extracts features from feature sparse text and improves the accuracy of text classification.(3)In the process of model training,on the one hand,Adam optimization algorithm is applied to accelerate the convergence speed of the model;on the other hand,Dropout technology is added to alleviate the over-fitting phenomenon of the model on small-scale data sets.(4)Based on the system framework and the integrated classification model,a hotel review text emotion analysis system is designed.The comment input module,data preprocessing module,comment classification module,comment query module and comment monitoring module are designed in detail.