Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Harnessing Deep LLM Participation for Robust Entity Linking

Jiajun Hou, Chenyu Zhang, Rui Meng

TL;DR

DeepEL presents a comprehensive framework that integrates Large Language Models across all stages of entity linking: candidate generation, disambiguation, and a global self-validation step that enforces sentence-wide coherence. By combining LLM-derived entity descriptions with traditional retrieval, formulating disambiguation as a controlled multiple-choice task, and applying a global validation using inter-entity context, DeepEL achieves robust improvements. Across ten benchmark datasets, it attains an average F1 increase of $2.6\%$ and a $4\%$ gain on out-of-domain data, without requiring fine-tuning. The work highlights the value of deep LLM integration for generalization and reliability in EL, while also noting inference instability in in-domain settings and outlining directions for further refinement.

Abstract

Entity Linking (EL), the task of mapping textual entity mentions to their corresponding entries in knowledge bases, constitutes a fundamental component of natural language understanding. Recent advancements in Large Language Models (LLMs) have demonstrated remarkable potential for enhancing EL performance. Prior research has leveraged LLMs to improve entity disambiguation and input representation, yielding significant gains in accuracy and robustness. However, these approaches typically apply LLMs to isolated stages of the EL task, failing to fully integrate their capabilities throughout the entire process. In this work, we introduce DeepEL, a comprehensive framework that incorporates LLMs into every stage of the entity linking task. Furthermore, we identify that disambiguating entities in isolation is insufficient for optimal performance. To address this limitation, we propose a novel self-validation mechanism that utilizes global contextual information, enabling LLMs to rectify their own predictions and better recognize cohesive relationships among entities within the same sentence. Extensive empirical evaluation across ten benchmark datasets demonstrates that DeepEL substantially outperforms existing state-of-the-art methods, achieving an average improvement of 2.6\% in overall F1 score and a remarkable 4% gain on out-of-domain datasets. These results underscore the efficacy of deep LLM integration in advancing the state-of-the-art in entity linking.

Harnessing Deep LLM Participation for Robust Entity Linking

TL;DR

DeepEL presents a comprehensive framework that integrates Large Language Models across all stages of entity linking: candidate generation, disambiguation, and a global self-validation step that enforces sentence-wide coherence. By combining LLM-derived entity descriptions with traditional retrieval, formulating disambiguation as a controlled multiple-choice task, and applying a global validation using inter-entity context, DeepEL achieves robust improvements. Across ten benchmark datasets, it attains an average F1 increase of and a gain on out-of-domain data, without requiring fine-tuning. The work highlights the value of deep LLM integration for generalization and reliability in EL, while also noting inference instability in in-domain settings and outlining directions for further refinement.

Abstract

Entity Linking (EL), the task of mapping textual entity mentions to their corresponding entries in knowledge bases, constitutes a fundamental component of natural language understanding. Recent advancements in Large Language Models (LLMs) have demonstrated remarkable potential for enhancing EL performance. Prior research has leveraged LLMs to improve entity disambiguation and input representation, yielding significant gains in accuracy and robustness. However, these approaches typically apply LLMs to isolated stages of the EL task, failing to fully integrate their capabilities throughout the entire process. In this work, we introduce DeepEL, a comprehensive framework that incorporates LLMs into every stage of the entity linking task. Furthermore, we identify that disambiguating entities in isolation is insufficient for optimal performance. To address this limitation, we propose a novel self-validation mechanism that utilizes global contextual information, enabling LLMs to rectify their own predictions and better recognize cohesive relationships among entities within the same sentence. Extensive empirical evaluation across ten benchmark datasets demonstrates that DeepEL substantially outperforms existing state-of-the-art methods, achieving an average improvement of 2.6\% in overall F1 score and a remarkable 4% gain on out-of-domain datasets. These results underscore the efficacy of deep LLM integration in advancing the state-of-the-art in entity linking.

Paper Structure

This paper contains 19 sections, 6 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Definition
  4. Methodology
  5. Entity Candidate Generation
  6. Entity Disambiguation
  7. Self-Validation
  8. Experiment
  9. Experimental Setup
  10. Datasets
  11. Baselines
  12. Evaluation Metric
  13. Main Result
  14. Different LLM Models in DeepEL
  15. Ablation Study
  16. ...and 4 more sections

Figures (1)

  • Figure 1: DeepEL's workflow. Text marked in red indicates the target entity mention currently being linked. Highlighted text represents already-linked entity mentions within the same sentence, providing contextual linking information for the target entity.