Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Prompt-based Graph Model for Joint Liberal Event Extraction and Event Schema Induction

Haochen Li, Di Geng

TL;DR

This work tackles the limitations of fixed event schemas in event extraction by introducing Liberal Event Extraction (LEE) and presents PGLEE, a prompt-based graph model that jointly extracts events and induces schemas without relying on external knowledge bases. The method generates candidate triggers and arguments through prefix-tuned prompts, encodes them in a heterogeneous event graph with trigger-argument and inter-trigger edges, and applies graph attention to produce semantic embeddings, which are then clustered and named to form event schemas. Empirical results on TAC-KBP 2017 show that PGLEE outperforms baselines when predefined schemas are scarce and can discover high-quality new event schemas such as Arrive and Collide, demonstrating effective schema induction and robust extraction. By removing dependence on predefined templates and external resources, PGLEE offers a scalable approach to cross-domain event extraction with automatic schema discovery and labeling capabilities.

Abstract

Events are essential components of speech and texts, describing the changes in the state of entities. The event extraction task aims to identify and classify events and find their participants according to event schemas. Manually predefined event schemas have limited coverage and are hard to migrate across domains. Therefore, the researchers propose Liberal Event Extraction (LEE), which aims to extract events and discover event schemas simultaneously. However, existing LEE models rely heavily on external language knowledge bases and require the manual development of numerous rules for noise removal and knowledge alignment, which is complex and laborious. To this end, we propose a Prompt-based Graph Model for Liberal Event Extraction (PGLEE). Specifically, we use a prompt-based model to obtain candidate triggers and arguments, and then build heterogeneous event graphs to encode the structures within and between events. Experimental results prove that our approach achieves excellent performance with or without predefined event schemas, while the automatically detected event schemas are proven high quality.

Prompt-based Graph Model for Joint Liberal Event Extraction and Event Schema Induction

TL;DR

This work tackles the limitations of fixed event schemas in event extraction by introducing Liberal Event Extraction (LEE) and presents PGLEE, a prompt-based graph model that jointly extracts events and induces schemas without relying on external knowledge bases. The method generates candidate triggers and arguments through prefix-tuned prompts, encodes them in a heterogeneous event graph with trigger-argument and inter-trigger edges, and applies graph attention to produce semantic embeddings, which are then clustered and named to form event schemas. Empirical results on TAC-KBP 2017 show that PGLEE outperforms baselines when predefined schemas are scarce and can discover high-quality new event schemas such as Arrive and Collide, demonstrating effective schema induction and robust extraction. By removing dependence on predefined templates and external resources, PGLEE offers a scalable approach to cross-domain event extraction with automatic schema discovery and labeling capabilities.

Abstract

Events are essential components of speech and texts, describing the changes in the state of entities. The event extraction task aims to identify and classify events and find their participants according to event schemas. Manually predefined event schemas have limited coverage and are hard to migrate across domains. Therefore, the researchers propose Liberal Event Extraction (LEE), which aims to extract events and discover event schemas simultaneously. However, existing LEE models rely heavily on external language knowledge bases and require the manual development of numerous rules for noise removal and knowledge alignment, which is complex and laborious. To this end, we propose a Prompt-based Graph Model for Liberal Event Extraction (PGLEE). Specifically, we use a prompt-based model to obtain candidate triggers and arguments, and then build heterogeneous event graphs to encode the structures within and between events. Experimental results prove that our approach achieves excellent performance with or without predefined event schemas, while the automatically detected event schemas are proven high quality.
Paper Structure (10 sections, 7 equations, 3 figures, 3 tables)

This paper contains 10 sections, 7 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Approach
  3. Event Candidates Generation
  4. Event Graph Construction
  5. Clustering and Event Schema Induction
  6. Experiments
  7. Experiment Setup
  8. Event Extraction for Predefined Types
  9. New Event Schema Induction
  10. Conclusion

Figures (3)

  • Figure 1: The target of Liberal Event Extraction.
  • Figure 2: The overall structure of PGLEE.
  • Figure 3: The Silhouette coefficient of different K selections.