"Catch Me If You Can" (CaMiCatzee) is a multimedia interlinking concept demonstrator. Our main goal is to show how multimedia assets (in our case still images) can be interlinked on the Web of Data. We envision to extend the User Contributed Interlinking (UCI) to multimedia assets. Clearly, the advantage is having high-quality semantic links from a multimedia asset to other data, hence allowing to connect to the linked datasets.
In flickr it is possible to annotate parts of a picture using so called "notes" (see also flickr's API). As the primary domain, we chose people depictions of people. Typically, flickr notes contain a string stating, e.g., "person X is depicted in this picture". However, there is no straight-forward way to relate this information with other data, such as FOAF data, locations, and contextual information (conference, holiday, etc.). This is where CaMiCatzee steps in: we apply the UCI principle by harnessing the fine-grained annotation capabilities of flickr in order to let people semantically annotate pictures.
CaMiCatzee's system architecture, showing the server as well as the client.
But CaMiCatzee offers more than pure interlinking: firstly, users are assisted in choosing a FOAF document (when URI or name has been selected, a sindice look-up suggests appropriate FOAF documents), and secondly in the "full report", the flickr tags of a picture are evaluated (again using sindice) and further used as a base for introducing rdfs:seeAlso links. This overview report is offered in XHTML+RDFa allowing consumption by both humans and machines.
This demonstrator is powered by FOAF, ARC2, and sindice semantic indexer; we would like to thank the people involved in building these technologies.