Modeling Task Interactions in Document-Level Joint Entity and Relation Extraction

TL;DR

This paper introduces Graph Compatibility (GC), a novel explicit interaction mechanism for document-level joint entity and relation extraction, improving performance by leveraging task interactions.

Contribution

It proposes Graph Compatibility to explicitly model task interactions, especially between coreference resolution and relation extraction, enhancing end-to-end document-level extraction.

Findings

GC achieves up to 2.3/5.1 F1 improvement over baselines.

Explicit task interaction modeling outperforms previous multi-task methods.

Experiments on DocRED and DWIE validate the effectiveness of GC.

Abstract

We target on the document-level relation extraction in an end-to-end setting, where the model needs to jointly perform mention extraction, coreference resolution (COREF) and relation extraction (RE) at once, and gets evaluated in an entity-centric way. Especially, we address the two-way interaction between COREF and RE that has not been the focus by previous work, and propose to introduce explicit interaction namely Graph Compatibility (GC) that is specifically designed to leverage task characteristics, bridging decisions of two tasks for direct task interference. Our experiments are conducted on DocRED and DWIE; in addition to GC, we implement and compare different multi-task settings commonly adopted in previous work, including pipeline, shared encoders, graph propagation, to examine the effectiveness of different interactions. The result shows that GC achieves the best performance by up to 2.3/5.1 F1 improvement over the baseline.