| In the age of networking and information-based,it is necessary to master the public opinion of network.There are up to 224 million users of Twitter in China in2020,so obviously to analyse the public opinion on Twitter has high research value and important social significance.On the public opinion analysis,there are some problems such as incomplete emotional dictionary and narrow domain usability about traditional text sentiment analysis,and not enough fine-grained and fewer classification methods(binary classification)on machine learning.Consequently,this paper proposes a method of sentiment classification to combine emotional dictionary and machine learning scheme,and to apply for the public opinion analysis of Twitter.The results of the experiment indicates that there is more accuracy than the traditional sentiment analysis,it is useful to master of trends in the public opinion and to make corresponding policy for government relevant departments.The first part of this paper is that to enlarge the emotion dictionary by building domain emotion dictionary and multidimensional emotion dictionary on concrete event in order to solve the problem of narrow domain usability,and to master correctly multifarious public opinion on the application of sentiment analysis.The second part of this thesis is that to collect key words on concrete event on machine learning cluster algorithm,especially to draw the main focus of public concern as the key words during the analysis of the fluctuation of emotion magnitude and polarity for the different time period by combining the first part work,which can be used to the public opinion analysis,and the application of the emotion analysis model named as “from emotion fluctuation to concern focus”.The third part of this paper is to design the software system of sentiment analysis,and to realize main functions,such as the visualization of the result of sentiment analysis,and to test this system in order to ensure the liability of system operation. |
PDF Full Text Request