MC^2: MPEG-7 Content Modelling Communities

Finding *specific* media content using major search engines, such as Google Video, and popular media sharing sites, such as YouTube, is difficult since the metadata searched against is too unstructured and general to reveal any specific results. Content models describe the features and semantics evoked to users by digital media at various levels of interpretation and enable more structured descriptions of content. Consequently, they have been the target of considerable research for some time. However, they require non-trivial computational and human resources to populate and maintain. With the rise of Web 2.0 and environments such as wikis, global communities of everyday users have begun to collaborate productively across the globe. What if the power of these communities could be harnessed, and the same principles applied, for the creation and maintenance of content models so that effort was distributed? Early, basic applications such as Flickr and del.icio.us that are based on simple community tagging suggest that this may be possible. However, how would these communities function with more advanced MPEG-7 content models and media streams? Could the complexity of MPEG-7 content models be simplified so that everyday users are able to work with them in this way? This project aims to pursue these research questions and attempts to build an online community-driven service which supports collaborative content modelling using the leading international metadata standard for media streams, MPEG-7. The usage patterns from users using the service will be collected, analysed and compared to those for community tagging and regular content modelling systems in order to demonstrate the effectiveness of the community approach.