Analysis And Research On Sentence Sentiment Orientation

With the2generation of the website, the provider of the information has not only be limited to the professors but also be open to the ordinary people. It enables the quick speed of the information. The increased information on the website daily is equal to the information in168millions of DVD, and that is more than a person can obtain in a whole life. There are3200million of reviews increased on the Facebook,200millions of twitters generated in Twitter every day. There are about tens of millions of transactions in Taobao everyday and the number has reached to517billions on the11th in November in2014. These numbers are the best proof of the rapid growth on the Internet.How to quickly find the information that people need in such large amount of information and how to quickly determine the sentiment orientation, which is useful in helping the users, companies and governments to make decisions, in it? This paper studies the sentence sentiment orientation and it can effectively help the people, comprises and governments when they need to make a choice.First, we make a brief description of the sentiment orientation analysis. We make clear of the boundary and the goal of the task. Then we introduce the main methods which prove to be helpful in solving the problem and be aware of the important parts of the task. The emphasis of the paper is the sentiment orientation of sentences. There are three technologies:based on emotion dictionary, based on syntactic relation and machine learning.Second, I propose a method of sentence sentiment orientation based on three-layers. It combines the emotion orientation and the syntactic relation of words. We divide the syntactic relations into three layers according to the distance. Meanwhile, we present the sentence in the form of a tree and make the sequence of the calculation considering the word relation and the tree structure.Third, we apply the method in Chinese text to get the final results. We observe a large amount of Chinese text and obtain the main structures in them. We divide the structures into three layers. Before the experiment, we clean the data, generate a emotion dictionary, identify the emotional sentences and make the model. Finally, we compare the results and get the weakness and strength of it.Besides, we implement the algorithm in English online reviews. We design the system which aims to analyze the reviews on English online website. After getting and analyzing the website, creating the general and feature dictionaries, obtaining the feature information and using the algorithm, we finally get the results.At last, I make a summary of the algorithm and point out the weakness that can be improved in future.