ULTRA: Unleash LLMs’ Potential for Event Argument Extraction through Hierarchical Modeling and Pair-wise Self-Refinement
Xinliang Frederick Zhang, Carter Blum, Temma Choji, Shalin D. Shah, Alakananda Vempala · 2024
Structural extraction of events within discourse is critical since it avails a deeper understanding of communication patterns and behavior trends.Event argument extraction (EAE), at the core of event-centric understanding, is the task of identifying role-specific text spans (i.e., arguments) for a given event.Document-level EAE (DocEAE) focuses on arguments that are scattered across an entire document.In this work, we explore open-source Large Language Models (LLMs) for DocEAE, and propose ULTRA, a hierarchical framework that extracts event arguments more cost-effectively.Further, it alleviates the positional bias issue intrinsic to LLMs.ULTRA sequentially reads text chunks of a document to generate a candidate argument set, upon which non-pertinent candidates are dropped through self-refinement.We introduce LEAFER to address the challenge LLMs face in locating the exact boundary of an argument.ULTRA outperforms strong baselines, including strong supervised models and ChatGPT, by 9.8% when evaluated by Exact Match (EM). * Work done during XFZ's internship at Bloomberg AI.News title: Drought puts 2.1 million Kenyans at risk of starvation News body: [0] National disaster declared as crops fail after poor rains and locusts, while ethnic conflicts add to crisis Last modified on Wed 15 Sep 2021 07.02 BST.[1] An estimated 2.1 million Kenyans face starvation due to a drought in half the country, which is affecting harvests.[2] The National Drought Management Authority (NDMA) said people living in 23 counties across the arid north, northeastern and coastal parts of the country will be in "urgent need" of food aid over the next six months, after poor rains between March and May this year .[3] The crisis has been compounded by Covid-19 and previous poor rains, it said, predicting the situation will get worse by the end of the year, as October to December rains are expected to be below normal levels.• • • [6] In July, the UN Food and Agriculture Organization in Kenya said the country needed 9.4bn Kenyan shillings (£62m) to mitigate the effects of the drought between July and November.• • • Event type: Droughts