Knowledge Discovery from Community-Contributed Multimedia

T he prevalence of imageand videocapturing devices and the advent of media-sharing services such as Flickr and YouTube have drastically increased the volume of community-contributed multimedia. For example, there are reportedly more than four billion images in Flickr and 24 hours of new videos are uploaded to YouTube every minute. Such a vast amount of photos, videos, and music shared via websites is bound to exert a profound social impact on human society and poses a new challenge for developing efficient indexing, search, mining, and visualization approaches for managing such large-scale media data. Social media is augmented with rich context, such as user-provided tags, comments, geolocations, time, device metadata, and so on. These bits of context data are promising resources to exploit for benefiting a wide variety of applications, such as annotation, recommendation, questioning and answering, advertising, and (cultural) activity discovery. The goal of this special issue is to present a concise reference of state-of-the-art efforts in such attempts for knowledge discovery over large-scale, community-contributed multimedia, and in particular the opportunities and challenges in this nascent arena. We have selected five articles that represent ways to exploit user-contributed photos and videos for several applications and that identify the theoretical challenges associated with managing such multimedia data.