Research On Opinion Leader Recognition And Sentiment Polarity Classification Of Microblog

In a few years, microblog develops explosively, through which public voices spread rapidly. The microblog opinion leaders play a very big role in the opinion guidance, and the contents of microblogs decide the final direction of public opinion. The technique of leader recognition and classification of sentiment polarity can be very effectively used in public opinion control. This research focuses on two relative techniques, the achievements are listed as follows:(1)For the existing weakness of recognition model in identifying opinion leaders and the lack of online opinion leader detecting method, a special topic based online opinion leader detecting method is proposed. It cost about a week to complete the online opinion leader detecting experiment under a certain topic. Experiment verifies that this method can not only collect the information of microblogs about a specified topic, but also find out the opinion leaders by calculate the influence of the topic-discussion participant dynamically.(2)Traditional text feature vector extraction method of n-gram tends to produce a high dimension, high-dimensional data not only increases the difficulty of classification, but also increases the time of data processing. To reduce the feature dimension, a feature extraction method based on Part-of-Speech tagging (POS) sequences is proposed. For the POS sequences can represent a kind of text, it is introduced to reduce the feature dimension here. This approach works well in microblog sentiment classification by extracting fewer features without sacrificing the classification accuracy. In order to verify the validity of the classification results of the method, two categories will be evaluated in the experiments.(3)In order to further refine the research mission of microblog sentiment polarity classification, this article proposed a classification idea based on microblog subspace, and puts forward an emotional word subspace, which can be used to improve the sentiment polarity classification based on dictionary. This method firstly divides microblog into different subspace of emotional word, and then classifies the microblog sentiment polarity by using the former method in corresponding subspace. The experimental comparison shows that the proposed method not only improves the classification accuracy, but also significantly improves the recall rate.