Research And Design Of Deceptive Network Information Filtering Platform For Public Opinion Monitoring

With the deepening degree of social information,the network public opinion and people's daily lives are becoming inseparable.In the face of a large number of online communities in our country,the organizations with ulterior motives hire the navy to focus on some socially sensitive issues and hot spots to anonymously publish many guiding ideas to guide the topics in the wrong direction.This is not conducive to social stability and development.Therefore,the use of computer technology to monitor the Internet public opinion has become a hot area and has a certain social value.This article explores the filtering of deceptive network information in key technologies of network public opinion monitoring system.As a major part of deceptive network information,deceptive opinions also have certain harm to society.So it is urgent to identify deceptive opinions.The deceptive opinion detection mainly extracts opinions content characteristics,and use machine learning to achieve the goal of detecting deceptive opinion.Most of the traditional machine learning is a shallow structure,so the complex text can not be represented.At the same time,deceptive opinions do not only reflect in the opinion text attribute,but also reflect in the commenter's behavior.So,from a single point of view of textual content attributes,only opinion text attributes can not fully consider the characteristics of deceptive opinions,which can all lead to loss of features.The paper uses the Convolution Neural Network(CNN)which is the one of depth learning framework to identify the content of deceptive opinion.The paper marks the online public hotel opinions to get a labelled data set.Then,the paper preprocesses the data,such as Chinese word segmentation and stop words,and uses the model Skip-Gram of tool word2vec to obtain the word vector.In the aspect of CNN model construction,the paper proposes the model which mixes text content attributes and behavioral attributes.Meanwhile,the traditional Convolution Neural Network is being improved from the point of the word order,to make convolution neural network more suitable for text classification.The experimental verification shows that the model proposed in this paper has achieved good results in the detection of deceptive opinions and can be used as an effective deceptive information filtering research method in practical work.