Few-Shot Document-Level Relation Extraction

TL;DR

This paper introduces FREDo, a new benchmark for few-shot document-level relation extraction, emphasizing the importance of document context and challenging the current models with realistic data and domain adaptation tasks.

Contribution

The paper presents a novel FSDLRE benchmark based on document-level data, adapting and improving the MNAV method for better domain adaptation, and highlighting the challenge of NOTA sampling.

Findings

FSDLRE is a challenging setting with unique characteristics.

Document-level corpora provide more realistic relation extraction scenarios.

The adapted MNAV method improves domain adaptation performance.

Abstract

We present FREDo, a few-shot document-level relation extraction (FSDLRE) benchmark. As opposed to existing benchmarks which are built on sentence-level relation extraction corpora, we argue that document-level corpora provide more realism, particularly regarding none-of-the-above (NOTA) distributions. Therefore, we propose a set of FSDLRE tasks and construct a benchmark based on two existing supervised learning data sets, DocRED and sciERC. We adapt the state-of-the-art sentence-level method MNAV to the document-level and develop it further for improved domain adaptation. We find FSDLRE to be a challenging setting with interesting new characteristics such as the ability to sample NOTA instances from the support set. The data, code, and trained models are available online (https://github.com/nicpopovic/FREDo).