Clustering of Videos on a Video Sharing Site Using User-Contributed Comments

Video sharing sites on the web offer a great amount of video contents, and some of them provide a facility for users to attach comments directly onto the video content. These comments let users know the content without seeing the video itself. In this paper, we propose a method for clustering the videos using the comments attached to them. We used the hybrid method of non-negative matrix factorization (NMF) and probabilistic latent semantic indexing (PLSI) for clustering. The method converts a higher dimension matrix into a lower one and finds topics in comments. Experimental results showed that the accuracy of the hybrid method was better than using only NMF.