LAAP: Learning the Argument of An Entity with Event Prompts for document-level event extraction

Document-level Event Extraction (DEE) aims to identify event types within a document and extract their corresponding arguments, which is essential for structured information provision in various NLP applications. Unlike sentence-level extraction, DEE requires handling events and arguments scattered across a document. Existing methods often focus on intricate feature interactions, neglecting explicit argument–entity relationships. We introduce a novel method, Learning the Argument of an Entity with Event Prompts (LAAP), which constructs event prompts for type detection, incorporating sentence placeholders to elicit event-specific information. Additionally, we propose an entity argument learning strategy that narrows down entity types to find the most suitable one. Experiments on the ChFinAnn and three other public datasets show that our method surpasses state-of-the-art approaches in accuracy and effectiveness.