Design And Implementation Of Internet Public Opinion Analysis System Based On Event Evolution And Sentiment Analysis

With the rapid development of mobile Internet technology,people can obtain and share information on the network more conveniently and quickly.The prosperity of social network platform gives people an opportunity to express their views and feelings freely on current political hot news events.However,the rise of the social network industry has virtually broken the law of the formation and diffusion of public opinion.Seemingly insignificant information fragments may be hyped by social media through the network,and then lead to a public opinion crisis.The formation speed and influence of network public opinion make both government and enterprises attach great importance to network public opinion and related technology.A variety of technologies have been proposed and applied in the emerging research field of network public opinion analysis.However,there is no unified standard for what kind of analysis results and services a network public opinion analysis system needs to provide to users.As a result,many researches are committed to improving the related technology of network public opinion analysis,but neglect how to present the analysis results in a way that is convenient for users to understand.Therefore,based on the combination of various network public opinion analysis related technologies,this paper mainly focuses on event evolution and emotion analysis to design and implement a network public opinion analysis system,and uses data visualization technology to display the analysis results.The system’s main functions are divided into data collection,data preprocessing,event evolution analysis,sentiment analysis,keyword extraction and information display etc.Data collection and data preprocessing provide sufficient and standardized data set for the system’s public opinion analysis.The event evolution analysis module adopts the CETM framework,comprehensively considers the co-occurrence dependency,event migration and temporal proximity of each news event in the news topic,and generates an event evolution diagram to help users sort out the evolution and development process of news topic.Sentiment analysis adopts the sentiment classification model based on deep learning.In order to solve the imbalance of the data set constructed from the Chinese text network data on the Weibo platform,based on the bidirectional LSTM with attention mechanism,the calculation of attention weight is improved by introducing emotional dictionary.Experiments show that the improved model effectively assigns attention weight to more important sentiment words,so as to achieve better performance in the sentiment classification of news comments on Weibo platform.The keyword extraction module combines the results of sentiment analysis in the process of keyword extraction of news topics,and fully considers the sentiment features of news comments.In addition,in order to make it easier for users to understand various abstract and complex public opinion analysis results,the system adopts various data visualization technologies,such as word cloud,stream graph,etc.,to display the analysis results more intuitively and clearly.An important topic of people’s livelihood,only by catching and guiding public opinion in time can we take precautions.The public opinion analysis system designed and implemented by this paper can make users clearly and conveniently understand public opinion on various news topics.