| With the rapid development of the Internet technology, microblog has become an important tool and platform for people to receive and deliver message. However, microblog text exists these problems, such as irregular grammar, colloquial language and network buzzwords. Moreover, most microblog text is short and contains little key information, which enable microblog public opinion analysis face certain difficulty and challenge.(1)microblog text is short, feature words are less, and it is hard to identify and express the semantics.(2)Based on the expressive methods of feature word vector, it exists the problems of high dimension and data sparseness.(3)Limited by the semantic expressive methods, the intelligent degree of microblog public opinion analysis software is low.Considering to the above problems existing in current microblog public opinion analysis, this thesis proposes the topic-based resolution on microblog public opinion analysis, and implements a microblog public opinion analysis prototype system. It mainly concludes:(1)Propose a method of microblog sensitive public opinion analysis, it learns the microblog topic with the aid of the LDA model, and constructs microblog topic vector and sensitive topic vector on the space of new feature words as well as identifying the microblog sensitive topic by computing its similarity. The experimental results show that this method can well identifies microblog sensitive topic.(2)Propose a method of microblog hotspot analysis. It learns the microblog topic with the aid of the LDA model, and identifies microblog hotspot public opinion by measuring how often the microblog topics appear in different documents of the corpus. The experimental results show that this method is feasible.(3)Design and implement a prototype system based on topic-dominated microblog public opinion analysis to realize the main functions, such as microblog data preprocessing, topic study, sensitive public opinion analysis, hot topic analysis and so on. |
PDF Full Text Request