Community-Contributed Media Collections: Knowledge at Our Fingertips - Community-Built Databases: Research and Development - page 22

new viewing patterns and novel forms of social interaction. Accordingly, an in-
creasing research effort has been devoted to analyzing and modeling user behavior
on social networking sites.
The abundance of contents generated by media-sharing communities could
potentially enable a comprehensive and deeper multimedia coverage of events.
Unfortunately, this potential is hindered by issues of relevance, findability, and
redundancy. Automated systems are largely incapable of understanding the seman-
tic content of multimedia resources (e.g., photos, videos, documents). Queries on
multimedia data are thus extensively dependent on metadata and information
provided by the users who upload the media content, for example, in the form of
tags. However, this information is often missing, ambiguous, inaccurate, or errone-
ous, which makes the task of querying and mining multimedia collections a nontriv-
ial one. Hence, user-contributed media collections present new opportunities
and novel challenges to mine large amounts of multimedia data and efficiently
extract knowledge useful to improve the access, the querying, and the exploration
of multimedia resources.
The aim of the chapter is to present how information retrieval and data mining
approaches can be used to extract and manage the content generated by media
communities. The chapter is organized as follows. Section 2.2 reviews different
collections of user-contributed media, such as YouTube, Flickr, and Wikipedia, by
presenting their main features. Section 2.3 discusses several aspects of the reviewed
media collections and their associated user communities. In particular, the structure
of the underlying social networks is analyzed. Different research efforts aimed at
studying media content distribution and user behavior are then presented. Finally,
the semantics of content found in these collections are addressed. Section 2.4
describes three taxonomies, one for each media collection, proposed to sum up
and classify the main research issues being addressed. An overview of results
achieved in the media annotation domain is presented in Sect. 2.5 , whereas
Sect. 2.6 describes diverse research efforts targeted at developing novel and
efficient data mining techniques to (a) extract relevant semantics from image
tags, (b) train concept-based classifiers and automatically organize a set of video
clips relative to a given event, and (c) efficiently categorize a huge amount of
documents exploiting Wikipedia knowledge. Finally, Sect. 2.7 draws conclusions
and suggests research directions offered by the community-built media collections.
2.2 Social Media-Sharing Communities
The past few years have witnessed the rapid proliferation of social networking sites,
wikis, blogs, and media-sharing communities. The advent of media-sharing com-
munities, partially spurred by the popularity of affordable hand-held image and
video recording devices (e.g., digital cameras, camera phones), has favored the
growth of social networks. Nowadays, hundreds of millions of Internet users are
self-publishing consumers. This has resulted in user-generated content (UGC)