Research On Sentiment Analysis For Chinese Microblog

Sentiment analysis refers to the use of natural language processing, text analysis and computational linguistics to identify and extract subjective information in source materials. After the rapid development in these ten years, it has been a hot research direction and has produced lots of commercial value and social value. Microblog is also growing rapidly in these years. Sentiment analysis research paid a close attention to this new potential field. Although the researchers have explored the applying of sentiment analysis for Chinese microblog, there still has much problem in this research area, such as the difficulty of marking the mass Chinese microblog corpus, the imbalance of sentiment class, the overtraining problem in training process and the useless of emotion object knowledge of Chinese microblog.This paper introduces the algorithm of active learning and Markov logic network for solving these problems that mentioned above. With the two algorithms, this paper does a further step of research on sentiment analysis for Chinese microblog.With the problem of the difficulty for marking Chinese corpus, this paper uses the active learning algorithm method. The active learning algorithm method can expand the scale of train data through the mass of unmarked corpus. The method is extremely suitable for solving this problem. For more step, this paper extracted text features that are appropriate for the sentiment analysis of Chinese microblog, proposed training degree threshold setup method, weight setup for iteration method and weight setup for imbalance of sentiment class method to optimize the sentiment analysis problem for Chinese microblog. The experiment of active learning method shows the effectiveness for algorithm and optimization method in the sentiment analysis research of Chinese microblog.The Chinese microblog corpus also contains a lot of emotion object knowledge. It is a valuable topic for researching the method for using knowledge in the sentiment analysis for Chinese microblog. This paper introduces the algorithm of Markov logic network to use the emotion object knowledge for this research. On the basis of traditional machine learning, Markov logic network algorithm can use the new knowledge easily with the first-order logic rules. With the using of emotion object knowledge, Markov logic network algorithm gets a big promote compare to the traditional machine learning method in sentiment analysis experiment for Chinese microblog.Furthermore, the lack of the unified platform for supervised machine learning makes it inefficient to research. The process of machine learning is similar at most time, so the paper has designed an effective and practice generalize supervised machine-learning platform to complete the research on sentiment analysis for Chinese microblog and more machine-learning mission. This platform has realized all the experiments that described in the paper. It has avoided duplication of effort and got a big promotion with the using of this supervised machine-learning platform for researching.