Abstract: ［Purpose/Significance］The emotional information contained in bulletscreen comments and subtitles of videos was extracted by using sentiment analysis technology and then visualized. Further analysis was made on the features and relationship of video contents and user emotion, which could provide reference for video content production and its retrieval approach. ［Design/Methodology］Models of sentiment analysis and comparison of bulletscreen comments and subtitles were built upon the Sentiment Dictionary. Moreover, emotional word extraction, emotional type partition, and emotional value calculation were also conducted, which could be used to study the changing tendency with the combinational use of time series. ［Findings/Conclusion］It has been found that: (1) the emotional intensity of bulletscreen comments is generally higher than that of subtitles; (2) the distribution laws of emotional categories of bulletscreen comments and subtitles are generally the same, but their proportions are different; (3) the emotional relationships between bulletscreen comments and subtitles could be classified into four categories. Similarities and differences in distribution of video bulletscreen comments and subtitles could be used to recognize the differences between the expressed emotional sentiments of video content and audience real emotions in a macroscopic manner, thereby evaluate the quality of videos. Using the emotional curve of bulletscreen comments and subtitles over time, the influential relationship between them two can be explored, and then assess the quality of a specific video segment from a microscopic viewpoint. ［Originality/Value］Sentiment analysis and comparison have been conducted for the content of videos and users' true feeling, which can be applied to find the emotional coincidence and difference between them, and provide new methods and tools for film quality evaluation, production, recommendation and other tasks.

Keywords: Bullet-screen comment, Subtitle, Sentiment analysis, Sentiment Dictionary, Video quality, Video recommendation