Using audio description for indexing moving images

This paper includes some of the results of a study that looks at three types of text for automatically deriving shot-level indexing to moving images. Audio description is a voice added to the sound track of moving pictures to provide information for the visually impaired. We analyse two one-hour parts of a television production broadcast as a mini-series in 1997. We compare our results with those of a previous study, which identifies some of the characteristics of audio description and the associated moving image. We found close correspondence among some aspects studied and for other aspects much less correspondence, but for reasons we are able to explain. In addition, in the process of conducting the current study we further developed our methodology and now feel that it is a mature method for analysing audio description text as a source for generating indexing to the associated moving image.