Cost-sensitive Regularization for Label Confusion-aware Event Detection

TL;DR

This paper introduces a cost-sensitive regularization approach for event detection that emphasizes confusing label pairs during training, leading to improved performance on TAC-KBP 2017 datasets in multiple languages.

Contribution

It proposes a novel cost-sensitive regularization method with estimators for measuring label confusion, enhancing event detection accuracy.

Findings

Significant performance improvements on TAC-KBP 2017 datasets

Effective handling of label confusion in event detection

Applicable to both English and Chinese datasets

Abstract

In supervised event detection, most of the mislabeling occurs between a small number of confusing type pairs, including trigger-NIL pairs and sibling sub-types of the same coarse type. To address this label confusion problem, this paper proposes cost-sensitive regularization, which can force the training procedure to concentrate more on optimizing confusing type pairs. Specifically, we introduce a cost-weighted term into the training loss, which penalizes more on mislabeling between confusing label pairs. Furthermore, we also propose two estimators which can effectively measure such label confusion based on instance-level or population-level statistics. Experiments on TAC-KBP 2017 datasets demonstrate that the proposed method can significantly improve the performances of different models in both English and Chinese event detection.