Research On Emotional Intensity Of Network Public Opinion Based On Semantic Perspective

With the rapid development of modern network process and various network media,network has already become a "double-edged sword".Not only does the Internet make it easier for Internet users to access hot information and opinions using blogs,BBS,WeChat,weibo and news groups etc.,but also the spread of extreme negative news events is becoming faster and faster.If the negative network public opinion is misguided,it will have a bad impact on social stability,harmony and healthy development,and even pose a serious threat to our cultural and political security.Therefore,the government and relevant departments should grasp the development of public opinion and improve the ability of monitoring the network public opinion,which is of great significance for maintaining the harmonious development of the society.The research direction of network public opinion includes the monitoring of sensitive words,the classification of public opinion,the trend of public opinion,and the analysis of emotion and so on.This paper selects the emotional analysis of network public opinion and analyzes the emotional intensity of a public opinion event.Emotional analysis can not only excavate netizens' attitude and emotional inclination to this event,but also improve the accuracy of information filtering,and it can also be used to predict the development trend of events.From the carrier perspective,the network public opinion text can be divided into news,blogs,forums and weibo texts.This paper selects weibo news as the network public opinion's text,and builds the emotion intensity model of the network public opinion based on the semantic angle.The specific research contents include the following aspects:(1)Study the construction method of Chinese sentiment dictionary.Mainly based on the HowNet semantic similarity method,and improve it in selecting the rules of benchmark words and computing the emotional weight algorithm.Drawing lessons from the Chinese ontology library to classify the intensity of emotional words,this paper also considers the choice of the emotional intensity in the selected reference.In the algorithm of calculating weight,we also add benchmark sentiment intensity as weight,and extend the emotional dictionary based on the benchmark to calculate the emotional intensity of the remaining words.(2)Study the extraction method of text keywords.Word co-occurrence is the method of extracting subject words and TF-IDF(Term Frequency and Inverse Document Frequency)is the basis of feature selection algorithm,Word co-occurrence extracts thematic features in short text significantly,while the effect of TF-IDCD(Term Frequency and Inverse Document Class Density)feature weighting algorithm is better than other feature weighting in stability and comprehensiveness.In this paper,we integrate and improve these two algorithms,and give them different weights,and merge the improved algorithm to extract the theme features in weibo text,which achieves good results.(3)Calculate emotional intensity in the text of network public opinion.Firstly,the preprocess text,and then construct the emotional intensity model of network public opinion based on the emotional dictionary.The effectiveness and accuracy of the emotional intensity model of network public opinion are verified by simulation experiments on the selected hot public opinion events.This paper has a certain reference and reference for the research of network public opinion early warning.