Towards visual words to words | IEEE Conference Publication | IEEE Xplore