Sentiment Classification Algorithm of Danmaku Comment Based on Modified Bayes Model
Ziyi Wang, Guanying Huang · 2021
To research the real time performance of Danmaku comments in sentiment classification, three clustering algorithms were used to divide comments into periods based on their time of occurrence. Research on these periods found some features of Danmaku comments: The outbreak of comments is of a periodical property; In the same period the comments have the same sentimental polarity. Based on above studies, a modified Bayesian algorithm based on period division is proposed. The algorithm uses the consistency of polarity in the same period to correct the Bayes model when processing sentiment classification. Experiments show that the new algorithm is effective and especially suitable for the Danmaku in high consistency of polarity.