Posted on December, 2020March, 2021LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding