Role Knowledge Prompting for Document-Level Event Argument Extraction

Document-level event argument extraction (DEAE) aims to identify the arguments corresponding roles of a given type in document. However, scattering and overlapping make DEAE face great challenges. In this paper, we propose novel model called Role Knowledge Prompting for Document-Level Event Argument Extraction (RKDE), which enhances interaction between templates through role knowledge guidance mechanism precisely prompt pretrained language models (PLMs) extraction. Specifically, it not only facilitates PLMs understand deep semantics but also generates all simultaneously. The experimental results show that our achieved decent performance on two public datasets, with 3.2% 1.4% F1 improvement Arg-C, some extent, addressed roles.