Research On Text Classification Method Based On Deep Learning

The meaning of the emergence of artificial intelligence is to let the machine process some tasks automatically,hoping that the processing results can reach or surpass the human level.At present,artificial intelligence technology has been implemented in education,finance,medical and other fields,and the application scenarios are expanding.However,there are still some deficiencies in many aspects of AI technology.As a sub field of artificial intelligence,natural language processing still has a major problem that the machine can not fully understand human language due to language diversity and semantic complexity.At the same time,it is also a common problem faced by many scholars in the industry.In this paper,we focus on the problems of insufficient semantic understanding and poor usability in the field of Chinese text classification in natural language processing.(1)Aiming at the problem that the existing common models are not accurate in dealing with the emotion classification of the actual specific field text,and considering the complexity of the actual text,this paper proposes an optimization algorithm.Based on the classic algorithm model Bert,this method optimizes the traditional text sentiment classification algorithm and the mainstream text sentiment classification algorithm.The algorithm adopts a three-stage learning method and uses vertical domain data for reinforcement learning,which greatly enhances the effect of semantic emotion understanding of the model.Compared with the traditional text classification model and the mainstream text classification model,the accuracy of the verification set of the algorithm is greatly improved.Finally,the model is applied to real life,and a simple case is completed.Finally,a good practical effect is achieved.(2)In view of the general effect of the mainstream text classification model Bert in dealing with novel text multi classification tasks,this paper presents an optimization algorithm for text classification based on xlnet model.This algorithm combines the advantages of autoregressive LM and autoencoder LM,uses xlnet to replace Bert,and adds confidence rules to optimize the text classification method of novel text data.Experimental results show that the accuracy of the algorithm is improved compared with the mainstream text classification methods.Through comparative experiments,it is proved that the accuracy of the model can be improved by adding confidence.To sum up,this paper mainly studies the text classification technology,using vertical domain data to enhance learning and add confidence rules to optimize the existing text classification algorithm,which provides some new ideas for the text classification optimization of the existing scheme.