A Language to Support Automatic Composition of Newscasts

Abstract

Video production involves the selection, manipulation, and composition of video segments to achieve a refined piece suitable for an intended audience. By associating metadata with each segment it is possible to automate this production process. For example, a mechanism is achievable for the purpose of creating dynamically assembled compositions for information customization applications including news-on-demand.

In this paper we propose a grammar and associated production constraints necessary to facilitate automatic video composition in the news domain. The grammar encompasses composition based on content as well as on structure of a newscast. In addition to providing a framework for logical composition of information, the grammar provides constraints for customization of information under bounds on the playout duration or the content selected by a user. We demonstrate how the language assists automatic information manipulation and composition of a newscast specifically when the data are acquired from various sources and delivered under limited resources.