Improving Long Tailed Document-Level Relation Extraction via Easy Relation Augmentation and Contrastive Learning
This addresses a critical issue for real-world information extraction scenarios where rare relations are often overlooked, though it is incremental as it builds on existing DocRE approaches.
The paper tackles the long-tailed distribution problem in document-level relation extraction (DocRE) by proposing an Easy Relation Augmentation (ERA) method and a contrastive learning framework (ERACL), which enhance performance on tailed relations and achieve competitive overall results compared to state-of-the-art methods.
Towards real-world information extraction scenario, research of relation extraction is advancing to document-level relation extraction(DocRE). Existing approaches for DocRE aim to extract relation by encoding various information sources in the long context by novel model architectures. However, the inherent long-tailed distribution problem of DocRE is overlooked by prior work. We argue that mitigating the long-tailed distribution problem is crucial for DocRE in the real-world scenario. Motivated by the long-tailed distribution problem, we propose an Easy Relation Augmentation(ERA) method for improving DocRE by enhancing the performance of tailed relations. In addition, we further propose a novel contrastive learning framework based on our ERA, i.e., ERACL, which can further improve the model performance on tailed relations and achieve competitive overall DocRE performance compared to the state-of-arts.