Recurrent event query decoder for document-level event extraction

Jing Kong,Zhouwang Yang

DOI: https://doi.org/10.1016/j.engappai.2024.108533

IF: 8

2024-05-13

Engineering Applications of Artificial Intelligence

Abstract:Document-level event extraction is a challenging task in natural language processing, as it involves multiple events within a document and scattered event arguments across sentences. To tackle these challenges, we propose a recurrent event query decoder, i.e., a recurrent module that dynamically updates event queries to capture cross-event dependencies. Our approach then generates arguments by extracting role-argument relations using bilinear mapping, which helps address the issue of scattered arguments. Experimental results demonstrate that our proposed approach outperforms state-of-the-art models on a large-scale public dataset and actual application data, achieving significant improvements in F1-score. In domain-specific event extraction applications, our method achieves higher accuracy with fewer resources compared to general-purpose large language models.

automation & control systems,computer science, artificial intelligence,engineering, electrical & electronic, multidisciplinary