Targeted Augmentation for Low-Resource Event Extraction

TL;DR

This paper proposes a targeted augmentation method with back validation to improve low-resource event extraction, producing more diverse and accurate training data, and demonstrating significant effectiveness in experiments.

Contribution

It introduces a novel targeted augmentation paradigm with back validation specifically designed for low-resource event extraction tasks.

Findings

Enhanced diversity, polarity, accuracy, and coherence in augmented data.

Significant improvements in event extraction performance.

Discussion of limitations and future directions.

Abstract

Addressing the challenge of low-resource information extraction remains an ongoing issue due to the inherent information scarcity within limited training examples. Existing data augmentation methods, considered potential solutions, struggle to strike a balance between weak augmentation (e.g., synonym augmentation) and drastic augmentation (e.g., conditional generation without proper guidance). This paper introduces a novel paradigm that employs targeted augmentation and back validation to produce augmented examples with enhanced diversity, polarity, accuracy, and coherence. Extensive experimental results demonstrate the effectiveness of the proposed paradigm. Furthermore, identified limitations are discussed, shedding light on areas for future improvement.