Combining multiple representations on the TRECVID search task [video retrieval system] | Publicación