| Social network sites provide the people with a multivariate open public opinion environment, the users can fully express their opinions. The sentiment analysis of text contents in social network is increasingly complicated, and it becomes more and more important under the trend of social network information explosion. Due to the existence of Chinese text semantic ambiguity and the strong relevance of semantic environment, the traditional English text analysis methods are too difficult to fit the complication and variety of Chinese text, so the analysis should be based on the characteristics of Chinese short text.At first, this paper makes good use of the semantic dictionary and extension rules of the phrase mode to complete the sentiment polarity classification. After formatting the short texts in social network, then we use the phrases matching method and emotional words statistics method to calculate the value of text tendentiousness, so as to realize the semantic sentiment analysis of short texts in the Internet. On the other hand, there are large amount of texts of social network in the different fields. Without semantic dictionaries of different fields, we take advantage of the existing datasets to extract the key features, use PCA method to reduce the sentence feature vector dimensions, and then utilize the neural network to realize fine-grained sentiment classification of short texts in microblog so as to grasp the sentimental information of short texts which are created by users.On the basic of the semantic dictionary and the extension rules of the phrase mode, through the analysis about the sentence structure of the short text, the improved algorithm has increased the accuracy of text sentiment analysis. And in order to realize the sentiment classification of social network short texts in the different fields, this paper proposes the multi-dimensional sentiment analysis method based on machine learning which provides a new idea for the fine-grained sentiment analysis. |
PDF Full Text Request