Analysis Of Emotional Tendency Of News In Xinjiang Violent Based On LSTM

In recent years,the terrorist attacks in Xinjiang have occurred,the rapid development ofnetwork technology,the openness of the network platform leads to the event of violence in the network communication process is easy to form a network of public opinion.Today’s hostile forces mainly via the network and the news media to guide the network or public opinion as the main channel for our penetration.Therefore,it is of great significance for us to take analysis timely and correctly of the emotional tendencies of the western media and the country,and to grasp the trend of social public opinion triggered by the violent events in Xinjiang.Firstly,this paper introduces the construction of Xinjiang violent corpus related technology,corpus construction process used to the concept of crowdsourcing,using semi-supervised labeling method to mark the data after the pre-treatment,and the use of a certain quality control means for corpus annotation Supervise and ensure the quality of the corpus.On the basis of the violent corpus in Xinjiang,this paper uses four different methods to complete the analysis of the emotional tendencies of the text.Using the traditional Word2 Vec and Doc2 Vec SVM method to carry on the text emotion tendency classification,at the same time,the depth learning method LSTM and its improved model TDLSTM model are used to classify the text,judge the emotional polarity of the news text,and then through the text emotion analysis result that analysis Western media and the country’s emotional tendencies in Xinjiang at a deeper level.Finally,we show the results of emotional analysis through visualization technology.