Building a Multi-Modal Thesaurus from Annotated Images | IEEE Conference Publication | IEEE Xplore