Improving Event Argument Extraction: A Machine Reading Comprehension-Infused Multi-Layer Label Pointer-Net

Event argument extraction, aimed at identifying entities and their associated roles within text containing pivotal events, constitutes a crucial step in structurally representing text. To study enhances the accuracy and efficiency of event argument extraction, especially when addressing challenges like entity and role overlap, we utilize a novel approach known as the Machine Reading Comprehension-infused Multi-layer Label Pointer-net for event parameter extraction(M2EE). Experimental results on the DuEE dataset show that M2EE outperforms traditional methods, particularly in F1 score and convergence speed. Additionally, M2EE surpasses previously assessed large language model (LLM)-based methods, showcasing its effectiveness in tackling intricate event parameter relationships.