TAP: Text-Aware Pre-training for Text-VQA and Text-Caption | IEEE Conference Publication | IEEE Xplore