Event Extraction in Large Language Model

TL;DR

The paper argues that event extraction should be reconceptualized as a system-level cognitive scaffold for LLM-centered pipelines, addressing reliability, traceability, and long-horizon memory. It surveys the evolution of EE from rule-based and classical learning to neural, generative, and instruction-driven paradigms, and extends the discussion to multimodal and cross-document settings. It catalogues tasks, decoding strategies, architectures, representations, datasets, and evaluation, then highlights open challenges and future directions for deployable, agent-ready EE in open worlds. The work emphasizes structured interfaces, graph-based retrieval, external memory, and neuro-symbolic verification as key directions for robust, scalable EE in the era of LLMs. Overall, it frames EE as a foundational perception and memory layer that enables reliable, grounded reasoning for open-world intelligent systems.

Abstract

Large language models (LLMs) and multimodal LLMs are changing event extraction (EE): prompting and generation can often produce structured outputs in zero shot or few shot settings. Yet LLM based pipelines face deployment gaps, including hallucinations under weak constraints, fragile temporal and causal linking over long contexts and across documents, and limited long horizon knowledge management within a bounded context window. We argue that EE should be viewed as a system component that provides a cognitive scaffold for LLM centered solutions. Event schemas and slot constraints create interfaces for grounding and verification; event centric structures act as controlled intermediate representations for stepwise reasoning; event links support relation aware retrieval with graph based RAG; and event stores offer updatable episodic and agent memory beyond the context window. This survey covers EE in text and multimodal settings, organizing tasks and taxonomy, tracing method evolution from rule based and neural models to instruction driven and generative frameworks, and summarizing formulations, decoding strategies, architectures, representations, datasets, and evaluation. We also review cross lingual, low resource, and domain specific settings, and highlight open challenges and future directions for reliable event centric systems. Finally, we outline open challenges and future directions that are central to the LLM era, aiming to evolve EE from static extraction into a structurally reliable, agent ready perception and memory layer for open world systems.

Figures (9)

• Figure 1: A roadmap of Event Extraction (1991–2025): Key milestones in the Pre-PLM, PLM, and LLM eras.
• Figure 2: Taxonomy of Event Extraction research in the LLM era, structured by tasks, modeling paradigms, learning strategies, and ecosystem.
• Figure 3: Task illustration for Text-based Event Extraction.
• Figure 4: Task illustration for different multimodal EE tasks.
• Figure 5: Different Methods for Event Extraction.