Look Again at the Syntax: Relational Graph Convolutional Network for Gendered Ambiguous Pronoun Resolution

TL;DR

This paper introduces an end-to-end gendered ambiguous pronoun resolver combining pre-trained BERT with Relational Graph Convolutional Network, significantly improving coreference resolution accuracy on the GAP dataset without fine-tuning BERT.

Contribution

It proposes a novel R-GCN based approach that leverages syntactic information to enhance BERT embeddings for coreference resolution, avoiding costly fine-tuning.

Findings

Achieved 80.3% F1 score on GAP dataset, surpassing previous 66.9%.

R-GCN embeddings outperform original BERT embeddings.

Method reduces computational cost by avoiding BERT fine-tuning.

Abstract

Gender bias has been found in existing coreference resolvers. In order to eliminate gender bias, a gender-balanced dataset Gendered Ambiguous Pronouns (GAP) has been released and the best baseline model achieves only 66.9% F1. Bidirectional Encoder Representations from Transformers (BERT) has broken several NLP task records and can be used on GAP dataset. However, fine-tune BERT on a specific task is computationally expensive. In this paper, we propose an end-to-end resolver by combining pre-trained BERT with Relational Graph Convolutional Network (R-GCN). R-GCN is used for digesting structural syntactic information and learning better task-specific embeddings. Empirical results demonstrate that, under explicit syntactic supervision and without the need to fine tune BERT, R-GCN's embeddings outperform the original BERT embeddings on the coreference task. Our work significantly improves the snippet-context baseline F1 score on GAP dataset from 66.9% to 80.3%. We participated in the 2019 GAP Coreference Shared Task, and our codes are available online.