Sentiment Polarity Analysis Of Microblog For Intelligent Security In Civil Aviation

With the continuous expansion of the scale of netizens,microblog meets people’s needs for hot information acquisition and free emotional expression.Microblog has become a crucial public opinion dissemination center and emotional expression platform.However,the rapid dissemination of information also makes it difficult to obtain timely and effective supervision of terrorist information.In recent years,terrorist information interference incidents emerge one after another in civil aviation industry.In order to prevent dangerous incidents in the civil aviation industry,this thesis proposes a risk level assessment method of microblog users for civil aviation intelligent security.This method combines with natural language processing technology,to monitor microblogs containing terrorist information that poses a security risk to civil aviation.Establish a database of key public opinion personnel by dividing the risk levels of microblog users.Increase the flight security intensity for key personnel.And assist security personnel to ensure civil aviation safety.First,collecting real data from microblog by crawler and then preprocessing.In this thesis,the experimental data adopts a large amount of real text data of microblog.Preprocessing can remove noise information such as repeated content,web page links and non-original content.These provide solid and reliable data support for subsequent research.Secondly,this thesis proposes a sentiment analysis model for mid-length microblogs based on capsule network.Since Microblog removed length limit,the average length of microblog text has increased year by year.However due to weak sentence correlation and high noise,the classification performance of mid-length microblog texts is poor.Therefore,in order to improve the detection effect of mid-length corpus.This thesis introduces attention mechanism.On the basis of fusing local and global features,using capsule vectors to extract deep emotional features.And it is proved that this model in this thesis has better performance and higher accuracy for mid-length microblog texts on the constructed microblogs data set.Finally,this thesis proposes a risk level assessment method of microblog users for intelligent security in civil aviation.This method collects microblog information related to civil aviation through keyword database search.After data preprocessing,subjective expression is obtained by screening.Then the risk level of this expression can be obtained through sentiment analysis.Additionally,collect historical data for key users,evaluate the risk level of the key users based on historical information.Establish a database of key public opinion personnel.At last,the effectiveness of the method is verified by simulation data.