Sentiment Analysis Of Chinese Micro Blog Based On Hybrid Model

With the popular of social networking,more and more users are acquainted with publishing their views on social networking platform.The rapid development of micro blog provides them a more capacious platform.Analyzing the emotion tendency of every user's comments in an event plays a significant role in many situations such as leading public opinion direction for public relations officials and question-answering system.Many researches had been studied about sentiment analysis by domestic and international scholars.However,Chinese micro blog comments have more difficulties than English because of no blank space separation and discrepancy of Chinese.There are all kinds of micro-blog platforms in China currently.We select Sina micro-blog as our data source because of its highest user penetration.We generate a layered sentiment analysis system of score first and classification second by training data from Sina's API,which is focused on the emotion features and emoticon in comment data.And building the training features.Then,we establish an ensemble classifier for processing positive and negative emotional comment data which is imbalanced by analyzing the percentage of emotion comments in all comments.All data set that has sentiment tendency is considered to be a category in the first layer.The other is considered to be a category.Positive and negative comments are considered in the second layer.Finally,data in COAE2014 become our test data.The result shows our emotion analysis system based on hybrid model get the accuracy of 92.15%in training data set and 84%in test dataset.And the result shows the validity of our system.