Studies On Sentiment Analysis Of Chapter Based On Rule And Statistics

Emotional orientation of text have great value in e-commerce, e-government and Public opinion analysis. How to give an accurate result of the expression is a target of emotional orientation of text. Feature extraction is one of the principal ways to improve the accuracy of the text orientation emotion. This paper made a study of feature extraction methods, which is based on chapter. The main contributions of this thesis are as follows:1) This paper use new method to analyses the Emotional orientation of chapter, give the related frameworks, analyses compatibility of the framework. We analyze the emotional orientation of sentence to get the result of the emotional orientation of chapter. The advantages of this method is that we can adjust the weight of the sentence according to location,the comments object and so on, and then we can get a more similar emotional density function。It can solve problems that cause by the traditional method,because the traditional method see the general chapter as a whole.2) The paper analyses the validity of the idea that merge the resource-based approach and statistical methods. And we use rules (resource methods) and decision tables (statistical methods) to verify the idea is feasible. The advantages of resource-based approach are that emotional vocabulary is accurate. But the shortcoming of this method is poor. But statistical methods are opposite. The idea merges resource-based approach and statistical methods into an integrated model that avoids some shortcomings caused by using them only.It is showed by the experiment that DLCR based on resource and statistical methods improve the accuracy of DL and CR obviously. The result of DLCR to analyze emotional orientation of chapter is better than SVM, Bayes. It indicates that DLCR is valuable in real world.