On learning multiple descriptions of a concept | IEEE Conference Publication | IEEE Xplore