With the rapid development of the Internet, micro blog is playing a more and more important role in people’s daily life. Micro blog is an information sharing, communication and acquisition platform based on the relationship between users in which users can access personal network via WEB, WAP and various clients and update message with 140 character maximum to share real-time information. On the platform, people can freely publish information, follow other users and hot social events and express their options and attitudes. Through the analysis of micro blog sentiment analysis, we can effectively mining the public’s attitude towards an event or a product which has significant commercial value. What’s more, micro blog sentiment analysis technology can also contribute to the development of other studies in natural language processing. Therefore, micro blog sentiment analysis has attracted an increasing attention of researchers and has gradually been becoming a research hotspot. This paper focuses on the key technical reasearch on Micro blog sentiment analysis, aiming at achieving efficient and accurate analysis of micro blog sentiment analysis by using information retrieval, data mining and machine learning combined with its own characteristics. The main contributionsof this paper can be concluded as follows: 1) Micro blog segmentation. Existing Chinese word segmentation tools can reach a high accuracy in traditional Chinese texts. However, it acts poorly in Micro blog text due to the characteristics of the Micro blog text itself. But the performance of the micro blog segmentation is important to the analysis. So in this paper, we propose a micro blog word segmentation method combining rules and maximum entropy model. Firstly, we pre-process the url, hashtag, emoticon and special symbols. Secondly, we use the maximum entropy combining the traditional features and external dictionary which is used to enhance the segmentation. Finally, we adopt a series of post-processing operations to get better result. Experiments show that the proposed method outperformes the state-of-the-art method significantly. 2) Sentiment analysis of micro blog based on rich features. Micro blog sentiment analysis has three main tasks: subjective and objective classification, positive and negative classification, and evaluation object extraction. This thesis mainly focuses on the first two tasks and proposes a Chinese Micro blog sentiment analysis method based on the feature dive...
修改评论