Harvesting Events from Multiple Sources: Towards a Cross-Document Event Extraction Paradigm

TL;DR

This paper introduces cross-document event extraction (CDEE), a new paradigm that integrates information from multiple sources to provide a comprehensive understanding of events, supported by a novel dataset and a multi-step extraction pipeline.

Contribution

It proposes the CDEE task, constructs the CLES dataset with extensive cross-document events, and develops a benchmark pipeline achieving 72% F1, advancing multi-source event extraction research.

Findings

Constructed the CLES dataset with over 70% cross-document events.

Developed a 5-step CDEE pipeline achieving 72% F1.

Highlights the complexity and potential of cross-document event extraction.

Abstract

Document-level event extraction aims to extract structured event information from unstructured text. However, a single document often contains limited event information and the roles of different event arguments may be biased due to the influence of the information source. This paper addresses the limitations of traditional document-level event extraction by proposing the task of cross-document event extraction (CDEE) to integrate event information from multiple documents and provide a comprehensive perspective on events. We construct a novel cross-document event extraction dataset, namely CLES, which contains 20,059 documents and 37,688 mention-level events, where over 70% of them are cross-document. To build a benchmark, we propose a CDEE pipeline that includes 5 steps, namely event extraction, coreference resolution, entity normalization, role normalization and entity-role resolution. Our CDEE pipeline achieves about 72% F1 in end-to-end cross-document event extraction, suggesting the challenge of this task. Our work builds a new line of information extraction research and will attract new research attention.