SVM-Based Sentiment Analysis Algorithm of Chinese Microblog Under Complex Sentence Pattern

With the development of Web2.0 era, as local information publishing and social networking platform of Twitter, microblog has become an important medium for people to share and propagate information. Sentiment classification for microblog has also become research hotspot in natural language processing field. By analyzing existing sentiment classification features and complex sentence patterns of microblog and directing at defects of current microblog sentiment classification in feature selection and extraction, this paper combined semantic relation between complex sentences and sentence features of complete sentence based on proposing features of sentence-level fine-grained embedding features and semantic features under complex sentence pattern so as to conduct effective analysis of microblog sentiment features under complex sentence context. It used SVM classification model to conduct comparative experiment, and results indicated that feature selection method proposed in this paper could improve performance of microblog sentiment analysis.