Research On Opinion Mining In Information Extraction

In recent years, Information has shown its significant strength in the tide of internet. The development of Social Network accelerates the appearance of Self-Media, which again gives a steady flow of the information tide. Then how to find out the right useful information becomes a big problem. Recently, Information Extraction (IE) becomes a good choice to solve the problem. Among all interesting directions of research of IE, Opinion Mining attracts lots of researchers. Opinion Mining mainly pays attention on the polarity of sentiment, the source of opinion as well as the target of opinion. Because of the fact that the sentiment information often show the like or dislike of users. The information related with sentiment often has more commercial value. Many transnational corporations rely on this kind of information to help with their ads systems and recommendation systems.Based on this background, this paper studies the development of research on Opinion Mining, verifies the performance of current classical information extraction methods used in Opinion Mining system. We also propose a hybrid model based on Ensemble Learning. Additionally, we do lots of experiments on the machine translation model, which is used in opinion target extraction. What’s more, we add the semantic information in the model to improve the performance. The main work of this paper is as the following:1. Based on the Ensemble Learning method, proposed a system using sequence labeling and rule-base method to extract the opinion factors.2. Research on the machine translation model in Opinion Mining and add semantic information to improve the performance. 3. Design and realize the "expert-system" used in opinion mining and verify its performance. Proposed a scalable Chinese New Word detection system using generalized suffix tree.