Multi-hop Evidence Retrieval for Cross-document Relation Extraction

TL;DR

This paper introduces MR.COD, a multi-hop evidence retrieval method that enhances cross-document relation extraction by effectively retrieving evidence across multiple documents, improving overall RE performance.

Contribution

The paper presents a novel multi-hop evidence retrieval approach with evidence path mining and a contextual dense retriever for cross-document RE.

Findings

Evidence retrieval with MR.COD improves cross-document RE accuracy.

MR.COD effectively retrieves evidence in both closed and open settings.

The proposed method outperforms existing retrieval techniques.

Abstract

Relation Extraction (RE) has been extended to cross-document scenarios because many relations are not simply described in a single document. This inevitably brings the challenge of efficient open-space evidence retrieval to support the inference of cross-document relations, along with the challenge of multi-hop reasoning on top of entities and evidence scattered in an open set of documents. To combat these challenges, we propose MR.COD (Multi-hop evidence retrieval for Cross-document relation extraction), which is a multi-hop evidence retrieval method based on evidence path mining and ranking. We explore multiple variants of retrievers to show evidence retrieval is essential in cross-document RE. We also propose a contextual dense retriever for this setting. Experiments on CodRED show that evidence retrieval with MR.COD effectively acquires crossdocument evidence and boosts end-to-end RE performance in both closed and open settings.