Analysis And Research On Public Sentiment Orientation For Social Security Incidents

The social media represented by micro-blog has developed into a new media,which brought convenience and freedom for people to obtain,publish and transmit information,and has completely changed people's information dissemination patterns and life patterns.However,the social media's characteristics including low cost,wide user and rapid spread have also brought convenience to the spread of social security incidents,enhanced harmfulness of social security incidents and made itself became an important position for foreign hostile forces and domestic lawless.These lawbreakers use the new media to spread rumors and publish cyber attacks,in order to achieve the goal of subverting state power,undermining national unity,impeding social stability and damaging the interests of the people.Therefore,the detection of social security incidents in social media,the analysis and prediction of public sentiment trends not only have important theoretical values,but also have important practical significance for maintaining national security and maintaining social stability.Taking Sina micro-blog as the research object,this paper conducts research on the detection method of social security incidents in micro-blog,the analysis method and prediction method of the sentiment tendency of the public.The research works cover the discovery,understanding,analysis and prediction of the Internet public opinion of the social security incidents,forming a set of methods and related theories of social security incidents' analysis.The main research works and achievements the following five aspects:(1)Research on the construction method of micro-blog basic resource baseA micro-blog data crawling method and a micro-blog text preprocessing strategy are proposed,and a micro-blog text preprocessing user dictionary is constructed.Through the research of micro-blog data crawling technology,a micro-blog data crawling strategy based on the combination of web crawler and Sina API is constructed.According to the characteristics of micro-blog text,a corresponding microblog text preprocessing scheme is formulated and a large-scale user dictionary is constructed,the dictionary includes more than 660,000 common word,more than 40,000 special term,and nearly 800 special term suffix.Completed the construction of the micro-blog basic resource database.(2)Research on calculation method of semantic relateness of wordsA word semantic relatedness calculation model based on semantic relationship graph is proposed.Through the extraction of semantic relations in HowNet semantic knowledge base and the extraction of semantic collocation relationship in large-scale corpus,the semantic relationship graph of words is constructed.Based on the semantic relationship graph,the related algorithms and the theories of graph theory,a word semantic relatedness calculation model is constructed.The experimental results show that the model has good performance in calculating the semantic relateness of words,and its domain adaptation can be improved by adding domain corpus.(3)Research on detection method of social security incidents in micro-blogA set of social security incidents feature word representation systems and social security incidents detection models in micro-blog are proposed.By making use of the emergency and topicality of social security incidents,the social security incidents feature word representation system is constructed based on five aspects including the basic elements of incident,the topic information of incident's,the labels information of incident,the emergency of incident,and the features of distinguishing incidents from advertisements.Combining the word semantic relateness calculation methods and the feature word representation system,a social security incident detection model based on hierarchical clustering and incremental clustering is constructe.Experiments show that the feature words extracted by the model have strong representativeness and the detection performance of social security incidents is better.(4)Research on the method of micro-blog text sentiment analysis and public sentiment tendency analysisA micro-blog text analysis model based on double attention mechanism and a public sentiment tendency analysis model are proposed.In view of the characteristics of micro-blog text sentiment expression,the micro-blog emotional dictionary,including six types of emotional symbol is constructed.Based on this,the attention mechanism is used to model the LSTM modeling results of micro-blog text and the emotional symbols contained in the micro-blog text respectively,and a micro-blog sentiment analysis model based on the double attention mechanism is constructed.Then,the tendency degree of sentiment tendency is used as a measure index,constructed a public sentiment tendency analysis model for specific social security incidents.Experiments show that the sentiment analysis model has a certain improvement over the existing best model.(5)Research on the method of public sentiment trend analysis and predictionA set of sentiment trend analysis indicators,and sentiment trend analysis and prediction models are proposed.In view of the demand for emotional trend analysis and prediction,a total of six sentiment trend analysis indicators,including the number of specific sentiment trends and the proportion of specific sentiment trends,are constructed.A sentiment trend analysis model is constructed based on sentiment trend analysis indicators and polynomial fitting.By merging the slope changes of the sentiment trend regression function under multigranular time slices,a sentiment trend prediction model was constructed.Through the analysis of relevant actual cases,it shows that the model has good performance.Finally,a public sentiment analysis system for micro-blog social security incidents is implemented based on SSM architecture,Bootstrap framework and Echarts components.The system consists of four modules: the acquisition and preprocessing of micro-blog data,the detection of social security incidents,the public sentiment analysis and the public sentiment trend analysis and prediction.It can perform experiments on the detection of social security incidents,sentiment analysis of micro-blog texts,and analysis and prediction of public sentiment trends.