Learning Global Features for Coreference Resolution

TL;DR

This paper introduces a method using RNNs to learn global entity-cluster representations for coreference resolution, improving performance especially on pronominal mentions without extra search.

Contribution

It proposes a novel approach to model global cluster information with RNNs, enhancing coreference systems beyond independent mention predictions.

Findings

Improved coreference resolution performance, especially on pronominal mentions.

State-of-the-art results achieved without additional search.

Effective integration of cluster-level features into end-to-end systems.

Abstract

There is compelling evidence that coreference prediction would benefit from modeling global information about entity-clusters. Yet, state-of-the-art performance can be achieved with systems treating each mention prediction independently, which we attribute to the inherent difficulty of crafting informative cluster-level features. We instead propose to use recurrent neural networks (RNNs) to learn latent, global representations of entity clusters directly from their mentions. We show that such representations are especially useful for the prediction of pronominal mentions, and can be incorporated into an end-to-end coreference system that outperforms the state of the art without requiring any additional search.