Recent advances of incorporating layout information, typically bounding box
coordinates, into pre-trained language models have achieved significant
performance in entity recognition from document images. Using coordinates can
easily model the absolute position of each token, but they might be sensitive
to manipulations in document images (e.g., shifting, rotation or scaling),
especially when the training data is limited in few-shot settings. In this
paper, we propose to further introduce the topological adjacency relationship
among the tokens, emphasizing their relative position information.
Specifically, we consider the tokens in the documents as nodes and formulate
the edges based on the topological heuristics from the k-nearest bounding
boxes. Such adjacency graphs are invariant to affine transformations including
shifting, rotations and scaling. We incorporate these graphs into the
pre-trained language model by adding graph neural network layers on top of the
language model embeddings, leading to a novel model LAGER. Extensive
experiments on two benchmark datasets show that LAGER significantly outperforms
strong baselines under different few-shot settings and also demonstrate better
robustness to manipulations.

该论文提出了一种新模型 LAGER，该模型基于文档中单词之间的拓扑相邻关系建立边缘，并通过引入图神经网络层将其结合到预训练语言模型中，从而有效地提高了实体识别的性能。