UniT: Multimodal Multitask Learning with a Unified Transformer | IEEE Conference Publication | IEEE Xplore