Abstract

In this paper, we propose a mechanism which acquires semantics of video contents as annotations from related Web communities. In particular, we have implemented a Web-based tool which user can easily create a Weblog entry quoting video scenes. This tool can acquire relationships which associate multiple video scenes with a document structure of a Weblog entry from editing histories. %In the result, we can acquire higher quality annotation than previous works which acquire information from communication with online message board systems. We have developed an online video quotation system ``Synvie.'' Moreover, we have analyzed real annotation data which were accumulated using the public beta service which we are providing, and confirmed the usefulness of our system. As an example of applications based on these annotations, we present a video scene retrieval system.