Personalized video summarization based on Multi-Layered Probabilistic Latent Semantic Analysis with shared topics

In this paper, we propose a multi-layered Probabilistic Latent Semantic Analysis (PLSA) model for personalized video summarization problem based on time synchronous comments offered by multiple users. Preliminary evaluations performed on an animation series of 624 minutes long with 12212 users show that the proposed model is able to capture the relationships among the preference of each individual user and the various video events, therefore is able to generate personalized summaries of unseen videos for different users.