Improving Generalization in Coreference Resolution via Adversarial Training

TL;DR

This paper introduces adversarial training to improve the generalization of coreference resolution systems, making them more robust to unseen named entities and enhancing their performance across datasets.

Contribution

It applies adversarial gradient-based training to a state-of-the-art coreference system, significantly boosting its ability to generalize to new named entities and datasets.

Findings

Improved performance on CoNLL dataset with unseen entities

Enhanced robustness to entity name variations

Higher accuracy on the GAP dataset

Abstract

In order for coreference resolution systems to be useful in practice, they must be able to generalize to new text. In this work, we demonstrate that the performance of the state-of-the-art system decreases when the names of PER and GPE named entities in the CoNLL dataset are changed to names that do not occur in the training set. We use the technique of adversarial gradient-based training to retrain the state-of-the-art system and demonstrate that the retrained system achieves higher performance on the CoNLL dataset (both with and without the change of named entities) and the GAP dataset.