Video Clip Recommendation Model by Sentiment Analysis of Time-Sync Comments

With the advent of video time-sync comments, users can not only comment the videos on the Internet, but also share their feelings with others. However, the number of the videos on the Internet is so huge that users do not have enough time and energy to watch all the videos. How to recommend the videos suitable for users has become an important problem. The traditional video sentiment analysis methods can not work effectively and the results are not easy to explain. In this paper, an emotion recognition algorithm based on sync-time comments is proposed, as a basis for the recommendation of video clips. First, we propose a formal description of video clips recommendation based on sentiment analysis. Secondly, by constructing the classification of time-sync comments based on Latent Dirichlet Allocation (LDA) topic model, we evaluate the emotion vector of the words in time-sync comments. Meanwhile, the video clips are recommended according to the emotion relationships among the video clips. The experimental results show that the proposed model is effective in analyzing the complex sentiment of different kinds of text information.