Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Nested Event Extraction upon Pivot Element Recogniton

Weicheng Ren, Zixuan Li, Xiaolong Jin, Long Bai, Miao Su, Yantao Liu, Saiping Guan, Jiafeng Guo, Xueqi Cheng

TL;DR

The paper tackles Nested Event Extraction (NEE) by focusing on Pivot Elements (PEs), which simultaneously serve as triggers for inner-nest events and as arguments for outer-nest events. It introduces PerNee, a five-module architecture that uses prompt-enhanced representations to improve trigger and argument modeling, with a dedicated Pivot Element Recognizer that converts PE recognition into trigger-pair relation classification, and a beam-search decoder for global structure optimization. To address data limitations in NEE, the authors present ACE2005-Nest, a generic-domain dataset built atop ACE2005 that expands nested-event types and roles. Empirical results show PerNee achieving state-of-the-art performance on ACE2005-Nest, Genia11, and Genia13, with ablations confirming the importance of PE recognition and prompt-based cues for robust NEE, and case studies illustrating practical gains in accurately recovering nested structures.

Abstract

Nested Event Extraction (NEE) aims to extract complex event structures where an event contains other events as its arguments recursively. Nested events involve a kind of Pivot Elements (PEs) that simultaneously act as arguments of outer-nest events and as triggers of inner-nest events, and thus connect them into nested structures. This special characteristic of PEs brings challenges to existing NEE methods, as they cannot well cope with the dual identities of PEs. Therefore, this paper proposes a new model, called PerNee, which extracts nested events mainly based on recognizing PEs. Specifically, PerNee first recognizes the triggers of both inner-nest and outer-nest events and further recognizes the PEs via classifying the relation type between trigger pairs. The model uses prompt learning to incorporate information from both event types and argument roles for better trigger and argument representations to improve NEE performance. Since existing NEE datasets (e.g., Genia11) are limited to specific domains and contain a narrow range of event types with nested structures, we systematically categorize nested events in the generic domain and construct a new NEE dataset, called ACE2005-Nest. Experimental results demonstrate that PerNee consistently achieves state-of-the-art performance on ACE2005-Nest, Genia11, and Genia13. The ACE2005-Nest dataset and the code of the PerNee model are available at https://github.com/waysonren/PerNee.

Nested Event Extraction upon Pivot Element Recogniton

TL;DR

The paper tackles Nested Event Extraction (NEE) by focusing on Pivot Elements (PEs), which simultaneously serve as triggers for inner-nest events and as arguments for outer-nest events. It introduces PerNee, a five-module architecture that uses prompt-enhanced representations to improve trigger and argument modeling, with a dedicated Pivot Element Recognizer that converts PE recognition into trigger-pair relation classification, and a beam-search decoder for global structure optimization. To address data limitations in NEE, the authors present ACE2005-Nest, a generic-domain dataset built atop ACE2005 that expands nested-event types and roles. Empirical results show PerNee achieving state-of-the-art performance on ACE2005-Nest, Genia11, and Genia13, with ablations confirming the importance of PE recognition and prompt-based cues for robust NEE, and case studies illustrating practical gains in accurately recovering nested structures.

Abstract

Nested Event Extraction (NEE) aims to extract complex event structures where an event contains other events as its arguments recursively. Nested events involve a kind of Pivot Elements (PEs) that simultaneously act as arguments of outer-nest events and as triggers of inner-nest events, and thus connect them into nested structures. This special characteristic of PEs brings challenges to existing NEE methods, as they cannot well cope with the dual identities of PEs. Therefore, this paper proposes a new model, called PerNee, which extracts nested events mainly based on recognizing PEs. Specifically, PerNee first recognizes the triggers of both inner-nest and outer-nest events and further recognizes the PEs via classifying the relation type between trigger pairs. The model uses prompt learning to incorporate information from both event types and argument roles for better trigger and argument representations to improve NEE performance. Since existing NEE datasets (e.g., Genia11) are limited to specific domains and contain a narrow range of event types with nested structures, we systematically categorize nested events in the generic domain and construct a new NEE dataset, called ACE2005-Nest. Experimental results demonstrate that PerNee consistently achieves state-of-the-art performance on ACE2005-Nest, Genia11, and Genia13. The ACE2005-Nest dataset and the code of the PerNee model are available at https://github.com/waysonren/PerNee.
Paper Structure (25 sections, 11 equations, 5 figures, 5 tables)

This paper contains 25 sections, 11 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Nested Event Extraction
  4. NEE Datasets
  5. Problem Formulation
  6. The PerNee Model
  7. The Text Encoder
  8. The Trigger Recognizer
  9. The Regular Argument Recognizer
  10. The Pivot Element Recognizer
  11. The Structure Decoder
  12. The ACE2005-Nest Dataset
  13. Nested Event Schema Discovery
  14. Data Analysis
  15. Experiments
  16. ...and 10 more sections

Figures (5)

  • Figure 1: Examples of flat (a) and nested (b) events.
  • Figure 2: The overall framework of the PerNee model.
  • Figure 3: Analysis of the proportions of event types capable of introducing nested events.
  • Figure 4: Detailed analysis on ACE2005-Nest.
  • Figure 5: Case study on the importance of PE recognition for NEE on ACE2005-Nest.