Visual Analysis Of Short Comment Text Based On Dual Time Dimension

Recently,the research on the analysis of massive danmaku data has mushroomed because of the prevalence of the danmaku culture.The danmaku data with two dimensions of temporal properties,a new commentary design for online videos,contains a host of valuable information,such as the sentiments,attitudes,and opinions of viewers.The information is more representative and comprehensive with the ever-increasing amount of danmaku data over time.However,extracting valuable and representative information from huge amounts of danmaku is a challenging task.Therefore,in this paper,an interactive visual analytics system is designed to visualize the sentiment and opinions of danmaku to aid in the analysis of video content and its impact.The main research elements of this paper are as follows.Firstly,for the characteristics of the danmaku data,the danmaku data is pre-processed,the sentiment values of the text fields are calculated and classified,the visualization results are used to continuously modify and expand the sentiment lexicon to improve the accuracy of emotion analysis,and the danmaku data is statistically analyzed from multiple dimensions for visualization purposes.Secondly,based on the cultural identity of the danmaku and the two-dimensional temporal properties of the danmaku,some suitable text visualization methods were adopted for visual analysis of the danmaku.The system uses color to encode the sentiment tendencies embedded in the graphical elements,visually extracts the danmaku information from four perspectives: the evolution of the distribution of the danmaku sentences over time,the distribution of the sentiment of the subject words in the video,the comparative analysis between the viewers and the development of the danmaku subject sentences,and provides a video view to assist users in inferring information.Furthermore,for different types of videos in the data set platform,certain rules are imposed on the data processing and visualization presentation to achieve flexible processing of different video lengths.Considering the large number of videos constantly generated and the constantly changing amount of danmaku of the same video,the system’s real-time processing of video danmaku is realized to enhance the practicality and convenience of the system.Finally,two types of videos,movies and short videos,are selected for case studies to explore the knowledge of data and reveal the phenomenon of the danmaku subculture.In addition,users of the system evaluate the system to verify its effectiveness and wide applicability and suggest future directions for the system’s improvement.