Human-Robot Dialogue Annotation for Multi-Modal Common Ground