Research On Text Classification Algorithm Based On Convolutional Neural Networks

With the development of information technology,internet data and resources are experiencing massive characteristics,and text information is exploding.How to quickly and accurately select the required data and information from the vast amount of data,text classification plays a very important role in the field of content information filtering and natural language processing.This thesis has done a lot of research and exploration based on CNN text classification model,proposed two practical and improved algorithms and carried out a lot of experimental analysis.Firstly,owing to the short length,sparse feature and strong context dependence of Internet short texts,a neural network model based on character-level embedded convolutional neural network and long-short-term memory network is proposed for short text classification.The model integrates the highway network framework to alleviate the difficulties in deep neural network training and improve the accuracy of classification.Through tests on several data sets,the results show that the proposed model is superior to traditional models and other CNN-based classification models in short text classification tasks.Secondly,aiming at the shortcomings of the classic CNN sentence classification model in dealing with long text tasks,and saving CNN's advantages in model parallelization,the LSTM gating mechanism is used to improve the layer-to-layer relationship in multi-layer neural networks.Further optimize the semantic representation of the text.A text classification algorithm based on gated convolutional networks is proposed.Experiments show that the algorithm can effectively improve its performance in text classification tasks on both Chinese and English data sets.Finally,based on the above research on character-level embedded text classification algorithm,this thesis further studies the application of the algorithm in cyberspace data management system.The system takes the event content,time,space and relationship in the narrative report as the research goal,and forms a cyberspace intelligence data management and analysis system supported by narrative technology.The text classification algorithm designed in this thesis has been applied in this system.