A Scalable Cross-Domain Event Extraction System via a Unified Generative Training Framework
2026-08-24 • Computation and Language
Computation and Language
AI summaryⓘ
The authors created a new method that uses one model to find events and their details in text all at once, instead of in separate steps. They trained this model on different kinds of event data so it works well across many topics and can handle lots of event types. They also made a website where users can upload documents, see the detected events and their parts, and compare results from different settings. This makes event extraction easier and more flexible for researchers and practitioners.
event extractionsequence-to-sequencepretrained language modelsevent detectionargument extractioncross-domain generalizationpipelineend-to-endinformation extractionvisualization
Authors
Siting Liang, Omar Adjali, Omair Shahzad Bhatti, Daniel Sonntag
Abstract
Event extraction is fundamental to information extraction. Prior approaches often separate event detection and argument extraction or depend on dataset-specific designs, limiting scalability and cross-domain generalization. We propose a unified generative sequence-to-sequence framework that performs event extraction subtasks jointly and supports both pipeline and end-to-end configurations. We fine-tune pretrained language models on multiple event datasets across diverse domains, enabling a single model to retain domain-specific semantics while generalizing over large and evolving label spaces. We demonstrate these capabilities through a web-based application tailored for researchers and practitioners. The platform supports document upload, schema-aware event extraction, visualization of triggers and arguments, and comparison of different extraction configurations across domains.