| In this paper, we propose | |
| the LayoutLM to jointly model interactions between text and layout information across scanned document images, which is | |
| beneficial for a great number of real-world document image understanding tasks such as information extraction from | |
| scanned documents. |