BERT for Coreference Resolution: Baselines and Analysis

TL;DR

This paper applies BERT to coreference resolution, demonstrating significant performance improvements and analyzing its strengths and limitations in understanding context and entity distinctions.

Contribution

It introduces BERT-based models for coreference resolution, providing baseline results and insights into model behavior and areas for future enhancement.

Findings

BERT-large outperforms ELMo and BERT-base in coreference tasks.

Achieved +3.9 F1 on OntoNotes and +11.5 F1 on GAP benchmarks.

Identified challenges in modeling document context and paraphrasing.

Abstract

We apply BERT to coreference resolution, achieving strong improvements on the OntoNotes (+3.9 F1) and GAP (+11.5 F1) benchmarks. A qualitative analysis of model predictions indicates that, compared to ELMo and BERT-base, BERT-large is particularly better at distinguishing between related but distinct entities (e.g., President and CEO). However, there is still room for improvement in modeling document-level context, conversations, and mention paraphrasing. Our code and models are publicly available.