Analysis Of Micro-blog Public Opinion In Sentiment Context

The rapid rises of mobile Internet is fundamentally changing the lives of human beings,people in the virtual society quickly and freely release messages,vent their emotions.More and more Internet users are accustomed to getting information in social media,resulting in cognitive,attitude and emotional inclinations of events.Micro-blog is one of the most representative social media platforms for public opinion dissemination.Its data covers all ages,industries and levels,and therefore it contains enormous social and commercial value.How to quickly and effectively conduct public opinion analysis on micro-blog information has become an urgent need at present.The purpose of this paper is to conduct a deep public opinion analysis of the micro-blog short texts,that is,quickly extract key information from the disordered microblog short texts and analyze their affective tendencies.The content of this paper is divided into the following aspects:(1)Keyword extraction of micro-blog short text.Combined with the changeable form of micro-blog short texts and the large amount of data,this paper uses the key words extraction technology based on TF-IDF algorithm and TextRank algorithm to extract key words from micro-blog texts.Then through the actual case analysis,we successfully extract the main information of public opinion events and the words of the key emotional attitude held by netizens to this event.So as to quickly obtain the emotional attitude of Internet users,discover problems in time,resolve conflicts and correctly guide the public opinion on the Internet.(2)Analysis of the public opinion of micro-blog short texts based on the sentiment dictionary.First of all,it introduces the mainstream sentiment dictionary in China at present,then integrates the mainstream Chinese sentiment dictionary,and adds a large number of words that cannot be identified into the sentiment polarity in micro-blog to the sentiment dictionary after judging the polarity of the sentiment word.Through the integration and expansion of existing sentiment dictionaries,we get the sentiment dictionary of the micro-blog of the network public opinion events,and analyze the hot events of the network public opinion based on the sentiment classification model of the sentiment lexicon to get the trend of sentiment and sentiment in a certain period of time.(3)Analysis of public opinion of micro-blog short text based on machine learning.In this paper,5 million micro-blog data shared by Zhang Huaping,Beijing Institute of Technology and NLPCC public dataset in 2013 and 2014 are used as word vector training corpus for Word2 Vec,and an sentiment classification model is constructed based on XGBoost algorithm.Based on the micro-blogging corpus,contextual relationships between words and words are fully considered,so as to effectively solve the neglect problem of word position and collocation and improve the accuracy of sentiment classification.