Hot Shot Detection And Recommendation Technology Of Bullet Subtitle Videos

In recent years,multimedia and streaming media technologies have developed rapidly,allowing users to watch online videos anytime and anywhere.When users choose which video to watch,they are accustomed to getting reference opinions from the video's historical comments or ratings.However,it is still difficult to find suitable video clips from those complicated and lengthy videos in a short period of time.Bullet subtitle is a new type of video interaction.The content of bullet subtitle is often highly related to the user's viewing experience and video content.Analyzing user sentiments in bullet subtitle videos plays an important role in improving the production of subsequent videos.Automatically detect hot clips in the video and push video clips that meet user preferences to users,which helps improve users' viewing experience and optimize the layout of video website resources.This thesis mainly takes BILIBILI video website as the research object and carries out research on the sentiment analysis of bullet subtitle video,detection and recommendation methods of bullet subtitle video hot clips.The main work is as follows.Firstly,aiming at the problem that the existing sentiment calculation algorithm is not fully applicable to the sentiment calculation of the bullet subtitle text,combined with the sentimental guidance function of the key bullet subtitle,a sentiment calculation algorithm combining key bullet subtitle and sentiment attenuation is proposed.At First,analyze and sort out the existing sentiment lexicon and bullet subtitle data to construct a Chinese bullet subtitle sentiment dictionary.Then,analyze the characteristics of key bullet subtitle,summarize the three elements of key bullet subtitle recognition,and identify the key bullet subtitle in the video clip.Finally,the sentiment attenuation model is introduced,and the sentimental decay and key bullet bullet subtitle sentimental guidance are integrated into the bullet subtitle sentimental calculation.The experimental results show that the proposed sentiment calculation method can effectively calculate bullet subtitle sentiments.Sencondly,aiming at the problem that the existing video hot segment detection methods are not suitable for the detection of bullet subtitle video hot segments,a video hot segment detection algorithm based on the sentiment and topic of bullet subtitle is proposed.At First,based on the change rate of the sentimental intensity of the video segment,the starting position of the hot segment is determined.Then,the Jaccard similarity algorithm is improved by combining word vectors and topic weights to calculate the topic similarity between video clips.Finally,based on the theme similarity of the video clips,the window range of the hot clips is determined.The experimental results show that the proposed detection method can effectively detect hot spots in the video,and the method has the best effect in practical applications.Thirdly,aiming at the problem that the data sparseness and the difficulty of feature extraction in current video recommendation algorithms,a video segment recommendation algorithm based on the sentiment and topic of the bullet subtitle is proposed.Combined with the barrage text to analyze the distribution of the video user's sentimental dimension and the user's preferences in terms of video topic and sentiment,and calculate the comprehensive similarity between video clips.Experimental results show that the algorithm can effectively recommend video clips based on the user's emotions and topic preferences.