Design And Implementation Of Sentiment Analysis System For Hot Microblog Comments

With the rapid development of mobile Internet,microblog has become the most convenient medium for people to get information and express opinions on the Internet.There are a lot of comments on social phenomena,current political news,economic hotspots and other events on microblog.The sentiment analysis of these comments is helpful to grasp the views and attitudes of social groups,to identify the impact of hot events in society more effectively,and to monitor public opinion and make decisions for management departments.In addition,the rational use of the sentiment analysis results of the hot microblog comments in different fields is also helpful to promote the development and upgrade the industry in the field.This thesis designs and implements a sentiment analysis system for hot microblog comments.The core part of the system is the sentiment analysis model of microblog text.In this thesis,a text correction model based on N-gram is adjusted in the method of data smoothing,which improves the accuracy of text correction on microblog data set.In the process of sentiment analysis of corrected data,this thesis proposes a new algorithm model called CGC(CNN-GRU-Cross).Then,this thesis combines text correction model with CGC to build the model TNSA(Text-correcting Method Based Neural Network for Text Sentiment Analysis),which can improve the accuracy of text expression and realize the parallel consideration of the semantic features and long-term dependence features of text data.This research shows the experimental results on two Chinese datasets.The results have confirmed that compared with the benchmark models,that the model proposed in this thesis can more accurately finish the tasks of microblog sentiment analysis.Firstly,this thesis summarizes the basic method of text sentiment analysis and the implementation technology of sentiment analysis system,then designs the sentiment analysis model in detail and verifies the effect of the model.Afterwards the thesis analyses the requirements of the sentiment analysis system,designs and implements the functional modules,and finally completes the functional tests and non-functional tests.The results show that the sentiment analysis system designed and implemented in this thesis has effectively completed the expected functions.