How To Tell Head From Tail in User-Generated Content Corpora


  • Nishanth Sastry King's College London



video-sharing, vimeo, long tail, geographical diversity, viral propagation of interest


This paper asks whether unpopular tail items in user-generated content corpora are important, and how tail items differ from the popular head items. We develop a user-centric characterisation of the tail which shows that although the head receives a disproportionate share of interest, tail items collectively serve a large number of users. "Tail seekers," with more like's in the tail than the head, are shown to constitute more than half the user base. We then examine how interests in head and tail items differ. Temporally, head items are found to enjoy a sustained interest, whereas interest in tail items is short lived. Spatially, interest in tail items is more geographically diverse. Finally, from a social angle, interest in unpopular items appears to be more "viral" than non-viral. We discuss implications of these observations for the handling and distribution of user-generated content.




How to Cite

Sastry, N. (2021). How To Tell Head From Tail in User-Generated Content Corpora. Proceedings of the International AAAI Conference on Web and Social Media, 6(1), 567-570.