Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Argument-Aware Approach To Event Linking

I-Hung Hsu, Zihan Xue, Nilay Pochh, Sahil Bansal, Premkumar Natarajan, Jayanth Srinivasa, Nanyun Peng

TL;DR

This paper tackles event linking by introducing an argument-aware framework that leverages explicit event arguments to disambiguate groundings and to detect out-of-KB instances. It augments input with event-argument tags and trains a cross-encoder with synthetic negative data to predict NIL, within a retrieve-and-rerank architecture that also includes a NIL candidate. Evaluated on Wikipedia (in-KB) and NYT (out-of-KB), the approach yields substantial gains, notably a 22% improvement in out-of-KB accuracy, and demonstrates robustness across verb/noun event mentions. The work highlights practical implications for robust event grounding and discusses future directions, including reducing tagging cost and exploring structured KB information, while acknowledging potential bias from using LLM-based data generation.

Abstract

Event linking connects event mentions in text with relevant nodes in a knowledge base (KB). Prior research in event linking has mainly borrowed methods from entity linking, overlooking the distinct features of events. Compared to the extensively explored entity linking task, events have more complex structures and can be more effectively distinguished by examining their associated arguments. Moreover, the information-rich nature of events leads to the scarcity of event KBs. This emphasizes the need for event linking models to identify and classify event mentions not in the KB as ``out-of-KB,'' an area that has received limited attention. In this work, we tackle these challenges by introducing an argument-aware approach. First, we improve event linking models by augmenting input text with tagged event argument information, facilitating the recognition of key information about event mentions. Subsequently, to help the model handle ``out-of-KB'' scenarios, we synthesize out-of-KB training examples from in-KB instances through controlled manipulation of event arguments. Our experiment across two test datasets showed significant enhancements in both in-KB and out-of-KB scenarios, with a notable 22% improvement in out-of-KB evaluations.

Argument-Aware Approach To Event Linking

TL;DR

This paper tackles event linking by introducing an argument-aware framework that leverages explicit event arguments to disambiguate groundings and to detect out-of-KB instances. It augments input with event-argument tags and trains a cross-encoder with synthetic negative data to predict NIL, within a retrieve-and-rerank architecture that also includes a NIL candidate. Evaluated on Wikipedia (in-KB) and NYT (out-of-KB), the approach yields substantial gains, notably a 22% improvement in out-of-KB accuracy, and demonstrates robustness across verb/noun event mentions. The work highlights practical implications for robust event grounding and discusses future directions, including reducing tagging cost and exploring structured KB information, while acknowledging potential bias from using LLM-based data generation.

Abstract

Event linking connects event mentions in text with relevant nodes in a knowledge base (KB). Prior research in event linking has mainly borrowed methods from entity linking, overlooking the distinct features of events. Compared to the extensively explored entity linking task, events have more complex structures and can be more effectively distinguished by examining their associated arguments. Moreover, the information-rich nature of events leads to the scarcity of event KBs. This emphasizes the need for event linking models to identify and classify event mentions not in the KB as ``out-of-KB,'' an area that has received limited attention. In this work, we tackle these challenges by introducing an argument-aware approach. First, we improve event linking models by augmenting input text with tagged event argument information, facilitating the recognition of key information about event mentions. Subsequently, to help the model handle ``out-of-KB'' scenarios, we synthesize out-of-KB training examples from in-KB instances through controlled manipulation of event arguments. Our experiment across two test datasets showed significant enhancements in both in-KB and out-of-KB scenarios, with a notable 22% improvement in out-of-KB evaluations.
Paper Structure (22 sections, 2 equations, 7 figures, 6 tables)

This paper contains 22 sections, 2 equations, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Statement
  4. Method
  5. Event Argument Tagging
  6. Negative Data Generation
  7. Model
  8. Experiments
  9. Experimental Settings
  10. Main Results
  11. Analysis
  12. Bi-Encoder Results
  13. Effectiveness of Negative Data Generation
  14. Conclusion
  15. Experiments with Large Language Models as Ranking Model
  16. ...and 7 more sections

Figures (7)

  • Figure 1: An example of two distinct events with the same event mention but different event arguments.
  • Figure 2: Given a text with an event to ground, our method extracts the event's attributes through event detection and argument extraction modules. The text, enriched with event argument tags, is then input into a Bi-Encoder to identify the top-$k$ potential nodes. These candidates are further ranked by a Cross-Encoder, which also considers an additional "NIL" candidate in case of out-of-KB instances. To equip the Cross-Encoder to robustly predict "NIL" and real KB entries, we train it with additional synthetic data generated through our negative data creation process.
  • Figure 3: Illustration for our negative data generation processing using LLM.
  • Figure 4: Prompt for our argument-aware data generation.
  • Figure 5: Prompt for non-argument-aware data generation baseline.
  • ...and 2 more figures