Research On The Application Of Image And Text Filtering Technology In Information Monitoring

With the rapid development of communication network and the rapid growth of intelligent terminal users,instant information release tools such as SMS,MMS,weibo,QQ and We Chat have been widely used.All of them have the advantages of convenient use and fast transmission.Among them,all the tools except SMS can carry both picture information and text information,so more applications can be made.However,due to the diversity and arbitrariness of information sources,the content it carries often contains bad pictures and text information.Therefore,content-based information filtering method must be adopted to identify,extract and analyze image and text content,so as to realize monitoring and filtering of bad information.Traditional information monitoring and analysis mode is usually implemented based on software automatic monitoring and manual audit,which has congenital defects in response speed,processing efficiency and labor cost.Modern automatic monitoring and analysis technology is mainly based on various machine learning algorithms,which can solve the problem of traditional mode better.However,faced with today's more complex mass information and specific application scenarios,they are often not satisfying in terms of cost and performance.In addition,the continuous improvement of computing environment and natural language processing technology has laid a good foundation for the in-depth research and application of information automatic detection and filtering technology.For this reason,this paper first carries on the in-depth study and analysis of the current commonly used image information and text information monitoring and analysis algorithms;On this basis,the two convolution Neural Network architecture VGG19 and ResNet50 in bad image content recognition performance analysis of the comparison and test validation,and choose better performance ResNet50 model is applied to the practical application platform;A text classification model based on BP neural network+word2vec is proposed to realize automatic monitoring and analysis of text information,and their validity is verified by using actual data.Finally,by taking the telecommunication mobile phone report information publishing system as a case,the paper realizes the practical application of the above research results and verifies its availability by analyzing and monitoring bad information.At present,the related application system has been put into practice and achieved good results of monitoring and analysis.