Research on network text sentiment classification algorithm based on multiple features

With the increasing trend of the number of the network text, the research on the sentiment classification of Web texts is becoming a hot issue. Based on the analysis and research of some existing classification methods, this paper proposes an optimization method based on the prior knowledge of the text topic. First, we pretreat the text of the corpus. Then the emotional polarity value of the text is calculated by the dictionary and semantic rules. Finally, according to the prior knowledge contained in the theme of the text and Naive Bayesian model, the text near the classification boundary will be reclassified. Experiment results show that this method is effective.