Key Technology Research For Group Sentiment Summarization For Microblog

With the rapid development of social network,more and more users are using the social network online platform, e.g., microblog, to communicate. There are some hot topics, having the close relationships among the social events people focusing on. Refining or generating the sentiment summarization of the microblog topics can mine the sentiment tendency conveniently and quickly. Furthermore, analyzing and mining the sentiment tendency of groups have significant roles for insighting the public opinions. Combining the relevant theory and knowledge of natural language processing and text summarization, this thesis proposes a group sentiment summarization approach for microblog. It analyzes the sentiment tendency and sentiment intensity of every relevant posts. On the basis of combining the text summarization technology, this thesis refines the summarizations with sentiment tendency. The main research works are as follows.1) In order to expand the scope of sentiment information processing, the research uses different sentiment dictionaries, and then assigns different weights for the relevant sentiment terms. Word2 Vec and semantic computation are used to improve the performance.2) Based on logistic regression model, a sentiment analysis model is proposed. This model analyzes the sentiment tendency and sentiment intensity by four-type sentiment element, i.e., sentiment dictionary, emoticons, sentiment phrases and negative sentiment words.3) Evaluating the candidates based on multi-dimensions model on the basis of analyzing the sentiment intensity, relevance and hot degree, etc. According to redundancy processing, the results having different sentiment tendencies are selected.Experiment shows that this proposed method can extract sentiment summarization of a topic effectively. The research contributes to know the main contents and sentiment tendencies of a topic in social network quickly, and the proposed approach has some values for scientific researchs and real applications. In the meantime, the deficiencies and the further works are also present in the end.