No bull, no spin

Abstract

User-contributed tags have shown promise as a means of indexing multimedia collections by harnessing the combined efforts and enthusiasm of online communities. But tags are only one way of describing multimedia items. In this study, I compare the characteristics of public tags with other forms of descriptive metadata'titles and narrative captions'that users have assigned to a collection of very similar images gathered from the photo-sharing service Flickr. The study shows that tags converge on different descriptions than the other forms of metadata do, and that narrative metadata may be more effective than tags for capturing certain aspects of images that may influence their subsequent retrieval and use. The study also examines how photographers use peoples' names to personalize the different types of metadata and how they tell stories across short sequences of images. The study results are then brought to bear on design recommendations for user tagging tools and automated tagging algorithms and on using photo sharing sites as de facto art and architecture resources.