Supporting the generation of argument structure within video sequences

Abstract

We describe our approach to the automatic generation of argument structures in the domain of video documentaries. Our approach releases control of the final video sequencing from the film maker/annotator to the system and thus allows users to select their own documentaries for viewing. Each video segment is annotated using a formal structure filled in with terms from a thesaurus. The annotations are used for finding and combining video segments into a final presentation. In order to influence the documentaries that can be generated, we introduce three methods for the annotator to evaluate the effectiveness of the annotations and to influence the process of automatic link generation.