Measurement Of Emotional Tendency And Analysis Of Influencing Factors In Haze Public Opinion

In recent years,China is in the process of rapid industrialization.The economic development h as made great achievements and the living standards of the residents have improved remarkably.However,with the rapid development of the economy,industrial agglomeration,energy consumption and overcapacity have become more prominent,and the air quality problem has become increasingly severe.The haze spreads across the country and has a significant negative impact on climate,environment,health and economy.And under the impetus of the development of the Internet,it has aroused widespread participation of the public.The public expressed their attitude towards the current situation of haze by commenting,forwarding and praising on the social platform.In this context,it is necessary to analyze the emotional characteristics of residents in the haze based on the microblog content published by users.On the one hand,it helps to grasp the opinions and appeals of residents on haze-related issues;on the other hand,it can timely discover the omissions in the social environment.At the same time,in order to correctly guide residents' attention and control the spread of negative emotions of residents in the haze,it is necessary to explore the macroscopic factors that lead to emotional differences in haze public opinion situation from a statistical perspective,and find out the direction of regulation and improvement.In this way,it can provide reference for timely adjustment of the negative emotions of residents in the haze incident and the development of reasonable early warning and response measures.Based on summarizing and analyzing the existing literature,Based on the relevant literatures at home and abroad,this paper conducts the research on the emotional tendency measurement and its influencing factors of haze through text mining,sentiment analysis and social network analysis methods.In the measure of sentiment orientation,firstly,based on text mining technology,this paper collects 690072 microblog data with the theme of haze in 2018,and performs noise processing,text segmentation and stop word processing,and retains 176076 microblog valid comments.Secondly,using the emotional vocabulary ontology library constructed by Dalian University of Technology and degree adverbs constructed with China Knowledge Network and negative lexicon to build a feature dictionary.And use the Word2Vec method and emoji word frequency statistics to extend the existing sentiment dictionary and extract the emotional information in the microblog text.Finally,based on the sentiment dictionary,this paper measures the emotional scores of residents in the haze and divides the emotion types.The accuracy,recall and F values are used to verify the accuracy of sentiment classification.Statistical visualization tools such as word cloud map,time trend line graph and thermal map are used to describe the emotional distribution from the perspective of type features and scoring features.The main conclusions are as follows:The main conclusions of the study are as follows:First,the negative emotions of the residents in haze public opinion are relatively high.Mainly manifested as:anger,sorrow,disappointment,fear and surprise about the current situation of haze;dissatisfaction and suspicion of the current situation of haze governance;anger against haze-related behaviors and fear of haze damage to physical and mental health.Second,in haze public opinion,the negative emotional intensity of the residents in 2018 showed a U-shaped trend in time distribution;in the spatial distribution characteristics,the spatial pattern of the decline from north to south was presented.Third,there are obvious spatial correlations between negative emotions in different provinces under the haze.According to the role of each location in the network structure,31 provinces are divided into four sections:northeast,northwest,southeast and southwest.The emotions in each section have significant correlations,and tend to converge under similar structural features,indicating that the negative emotions of different regions in haze public opinion are more obvious.Fourth,the inter-provincial air quality,pollution control,income level,Internet development and spatial distance have a significant negative impact on the emotional association of residents under the haze.That is to say,the smaller the difference between the indicators,the more similar the emotional intensity of the residents.Among them,the air quality situation and the pollution control intensity standardization coefficient are relatively high,and the residents' emotional influence in haze public opinion is greater.The innovation of this paper mainly includes the following three aspects:First,this paper collects 690072 texts of haze-themed microblogs published by residents of 31 provinces(excluding Hong Kong,Macao and Taiwan)in 2018,as the basic data source for haze emotion research.Secondly,this paper uses social network analysis to more accurately examine the statistical characteristics of spatial associations of negative emotions among provinces.It has the characteristics of global analysis and avoids the limitations of "adjacent" or "similar".Third,the current research on the influencing factors of emotions considers the individual microscopic characteristics such as gender,age,and education.However,The problem of haze is a macro-social problem,and the macro-factor has a great influence on the residents' feelings in haze public opinion.Therefore,this paper explores the influencing factors of residents' emotions in the haze from the macroscopic perspective,and explores the factors that lead to emotional differences.In order to timely adjust the negative emotions of residents in the haze incident,and provide reasonable early warning and response measures to provide reference.The shortcomings of this paper are mainly reflected in two aspects:First,in the study of sentiment orientation based on sentiment lexicon,the comprehensiveness of vocabulary in sentiment lexicon is the main problem of research.In this paper,the Word2Vec word vector training method is used to extend the sentiment dictionary based on the existing emotional vocabulary ontology library,but it still cannot cover all the emotional words in microblog content.Therefore,it is necessary to further improve the emotional dictionary.Second,there are some irony in the residents'publication of Microblog content,which will lead to a certain degree of bias in the research results.