Reasoning about video: knowledge-based transcription and presentation

Growing use of video data in a wide range of computer applications testifies to its high bandwidth and rich representational power but experience brings to light its limitations. The authors show how existing video transcription and presentation methods fail to make full use of the potential of video. Concerning transcription, they identify the semantic unpredictability problem as one of not knowing a priori the intervals in the video record that will be required for subsequent presentation. The syntactic ambiguity problem stems from inconsistent use of labels by video annotators and by automatic logging agents. Concerning presentation, they recognize that traditional models do not permit run-time generation of user-tailored output. The knowledge-based framework called intent-based authoring addresses the above problems to provide savings in human effort in both the annotation and presentation phases of video authoring.<>