LaTr: Layout-Aware Transformer for Scene-Text VQA | IEEE Conference Publication | IEEE Xplore