Year: 2026
Type: conference-paper
Abstract: We present a multi-agent reflective architecture for event extraction based on generative large language models (LLMs).Our architecture is the first of its kind to perform Semi-Open Event Extraction (SOEE), a hybrid framework that combines a fixed set of eventtemplate fields with dynamically generated attributes produced using self-reflection.A further contribution of the system is its operationalization of reflection as an internal question-generation and answering process.It is defined as the generation of questions about missing or implicit event information and finding their answers within the system itself.We model event extraction as an iterative dialogue between a reflective LLM-based agent, which generates questions to uncover missing event information and a set of expert agents, which provide domain-aware answers to these questions.The expert agents also generate the initial event template using a generative LLM.Across all evaluation experiments on articles from the health domain, MAREA demonstrates strong core-field extraction and effective reflective template expansion, with its three questiongeneration strategies producing useful additional event attributes.The observed errors were mainly due to imprecise prompt interpretation or inaccurate interpretation of the source article by the LLM, rather than to hallucination or an intrinsic inability to retrieve the requested information.
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166853464