Rewarding Coreference Resolvers for Being Consistent with World Knowledge

TL;DR

This paper proposes a method to improve coreference resolvers by rewarding their outputs based on the accuracy of extracted knowledge triples against knowledge bases, using multi-task reinforcement learning.

Contribution

It introduces a novel reinforcement learning approach that aligns coreference resolution with knowledge extraction accuracy, enhancing resolver performance.

Findings

Improved coreference resolver performance over state-of-the-art methods.

Effective use of multi-task reinforcement learning for resolver training.

Enhanced relation extraction accuracy through better coreference resolution.

Abstract

Unresolved coreference is a bottleneck for relation extraction, and high-quality coreference resolvers may produce an output that makes it a lot easier to extract knowledge triples. We show how to improve coreference resolvers by forwarding their input to a relation extraction system and reward the resolvers for producing triples that are found in knowledge bases. Since relation extraction systems can rely on different forms of supervision and be biased in different ways, we obtain the best performance, improving over the state of the art, using multi-task reinforcement learning.