Text this: A Qualitative approach to learning of spatial and temporal relations for cognitive computer vision systems