Research And Implementation Of Online News Text Classification System Based On Deep Learning

With the rapid development of Internet technology,text data show massive characteristics.Online media and new media platforms become an important part of news dissemination,and online news becomes one of the important sources of information for people.In order to meet the needs of a large number of online news readers and to improve the efficiency of personalized recommendation of content distribution platform,it is urgent to manage and utilize online news effectively.Based on this,this thesis focuses on the research and implementation of online news text classification.The goal is to build a text classification model for online news with higher accuracy.The work done in this thesis mainly includes the following parts:(1)Based on the short length of online news text and the close relationship between keywords and category tags,this thesis proposes a novel attention mechanism called the "full attention mechanism".Different from the traditional attention mechanism,full attention mechanism allocates attention to previous steps and current step at each step,so it can make the resource allocation more reasonable,make key information play different roles at different steps,and make the information discarded by the encoder be reused.(2)This thesis applies full attention mechanism to convolutional neural network and Bi-GRU respectively,and proposes FACNN and FABG.In order to verify the effectiveness of our work,experiments are carried out in English dataset(agnews)and Chinese dataset(chnews).The results show that the accuracy of FACNN and FABG is higher than that of other models,which verifies the effectiveness of full attention mechanism,FACNN and FABG.The accuracy of FABF in Agnews is 91.79%,which is 1.15%higher than that of Bi-GRU with attention.(3)In order to verify the effectiveness and stability of the model in practical applications,this thesis builds an online news text classification system based on the proposed model.The system provides a front-end interface to interact with users,and integrates FACNN and FABG in the background to return results.In this system,users can select the language of the text to be classified,enter the text,and obtain the final classification result.