GEGA: Graph Convolutional Networks and Evidence Retrieval Guided Attention for Enhanced Document-level Relation Extraction

Document-level relation extraction (DocRE) aims to extract relations between entities from unstructured document text. Currently, some studies are utilizing logical rules within evidence sentences to enhance the performance of DocRE. However, in cases where the data does not provide specific evidence sentences, researchers must generate a list of evidence sentences for the entire document through evidence retrieval (ER). This process typically faces a challenge, which is that the relevance between evidence and entity pairs is often weak. Moreover, another challenge in DocRE is the insufficient extraction of complex cross-relations between long-distance entities, which is often due to an inadequate understanding of long-distance semantic text. To overcome these challenges, we propose GEGA, a novel model for DocRE. The model leverages graph neural networks to construct multiple weight matrices, guiding attention allocation to evidence sentences. It also employs multi-scale representation aggregation to enhance ER. Subsequently, we integrate the most effective evidence information and semantic information to implement both fully supervised and weakly supervised training processes for the model. We evaluate the GEGA model on two widely used benchmark datasets. The experimental results indicate that our model has achieved comprehensive improvements compared to other existing SOTA models.