Towards collaborative annotation for video accessibility