Moving on from OntoNotes: Coreference Resolution Model Transfer

TL;DR

This paper investigates how neural coreference resolution models trained on OntoNotes transfer to other datasets, emphasizing the importance of continued training especially with limited target data, and establishes new benchmarks including state-of-the-art on PreCo.

Contribution

It systematically evaluates coref model transferability across multiple datasets and introduces new benchmarks, highlighting continued training benefits.

Findings

Continued training improves transfer performance.

Transfer effectiveness is higher with fewer target documents.

Achieves state-of-the-art on PreCo dataset.

Abstract

Academic neural models for coreference resolution (coref) are typically trained on a single dataset, OntoNotes, and model improvements are benchmarked on that same dataset. However, real-world applications of coref depend on the annotation guidelines and the domain of the target dataset, which often differ from those of OntoNotes. We aim to quantify transferability of coref models based on the number of annotated documents available in the target dataset. We examine eleven target datasets and find that continued training is consistently effective and especially beneficial when there are few target documents. We establish new benchmarks across several datasets, including state-of-the-art results on PreCo.