PoKE: A Prompt-based Knowledge Eliciting Approach for Event Argument Extraction

TL;DR

PoKE introduces a prompt-based method for event argument extraction that effectively elicits structured knowledge from pre-trained language models, outperforming existing methods especially in low-resource settings.

Contribution

The paper proposes a novel prompt-based approach for event argument extraction that captures both independent and joint event knowledge, improving performance over prior methods.

Findings

Outperforms recent methods on ACE2005 dataset

Effective in both fully-supervised and low-resource scenarios

Demonstrates advantages of prompt-based knowledge elicitation

Abstract

Eliciting knowledge from pre-trained language models via prompt-based learning has shown great potential in many natural language processing tasks. Whereas, the applications for more complex tasks such as event extraction are less studied since the design of prompt is not straightforward for the structured event containing various triggers and arguments. % Meanwhile, current conditional generation methods employ large encoder-decoder models, which are costly to train and serve. In this paper, we present a novel prompt-based approach, which elicits both the independent and joint knowledge about different events for event argument extraction. The experimental results on the benchmark ACE2005 dataset show the great advantages of our proposed approach. In particular, our approach is superior to the recent advanced methods in both fully-supervised and low-resource scenarios.