Video sharing websites study Content characteristic analysis

In this paper we present a recent study on video sharing websites. This study aims to understand their content characteristics. This could be useful to understand Internet users´ behaviour and manage web resources in order to provide a better video sharing service. In our work, we improved an existing graph-sampling algorithm so that it could be more adapted to sample over the video sharing websites. We crawled over 13 millions videos on YouTube and DailyMotion. We re-classified YouTube and DailyMotion content with our new category system and analysed the content category distribution and popularity of these two websites. We find that content in the “Media” category takes a large proportion in both websites, and also that the content category popularity does not depend on its proportion. Besides we then analyse the video duration and figure out that most videos on the video sharing websites are short, within several minutes. We study video count of views as well and find that the distribution of video count of views can be approximated by a negative exponential distribution that is long-tailed. That is to say, most of videos have a small or medium count of views; only a few videos can have a count of views of a bigger order of magnitude.