A comparative study of spatial, temporal and content-based patterns emerging in YouTube and Flickr

Due to the recent advances and wide adoption of Web 2.0 technologies, there is an abundance of publicly available user generated content, which can be a valuable resource for researchers, enabling them to apply sophisticated analysis methods on data of unprecedented scale. This paper focuses on the detection and comparison of spatial, temporal and content-based patterns from user-generated content available through two major content-sharing services YouTube and Flickr aiming to explain some of the observed differences. In particular, we use different clustering approaches and visualization techniques to discover interesting spots in a city using the publicly available content (images, videos and associated metadata), we then classify them as either landmarks or events, and compare their ranking for each of the respective service. Based on that ranking, we confirm the expected pattern that people tend to take pictures of still objects (monuments, sights) and make videos of events. We also consider the temporal aspect of the data, and extract movement trajectories of users. Lastly, we discover that even single users can generate noticable patterns, and that people find diverse uses of these content- sharing services.