A system for customized news delivery from video archives

Abstract

Video is a powerful medium for disseminating news as information. Like any other information, techniques are required to help search and locate interesting video content. We consider the construction of a system to automatically assemble an individualized news video when a sufficient volume of video is available for composition. We propose a vocabulary and representation for composition of news video. A model is presented that is used to store metadata and customize newscasts. The model facilitates representation of information clustering, object stratification, concept hierarchy, temporal ordering, and news "tours". The proposed customized news delivery system attempts to achieve a level of composition found in existing television news delivery.