Advancing Cross-Lingual Entity Alignment with Large Language Models: Tailored Sample Segmentation and Zero-Shot Prompts

Linyan Yang, Jingwei Cheng, Fu Zhang


Abstract
In recent years, the advent of large language models (LLMs) like GPT and Llama has significantly influenced numerous domains, particularly in advancing natural language processing (NLP) capabilities. LLMs have shown remarkable performance in NLP tasks such as relation extraction (RE) and knowledge graph completion (KGC), enhancing activities related to knowledge graphs. As a result, there is a growing interest in integrating LLMs into cross-lingual entity alignment (EA) task, which aims to identify equivalent entities across various knowledge graphs, thereby improving the performance of current baselines. However, employing LLMs for entity alignment poses challenges in efficiently handling large-scale data, generating suitable data samples, and adapting prompts for the EA task. To tackle these challenges, we propose Seg-Align, an innovative framework that integrating distance feature extraction, sample **Seg**mentation, and zero-shot prompts. Through extensive experiments on two widely used cross-lingual benchmark datasets, we have not only demonstrated the effectiveness of our proposed sample segmentation algorithm but also highlighted the state-of-the-art performance of Seg-Align. Code is available at https://github.com/yangxiaoxiaoly/Seg-Align.
Anthology ID:
2024.findings-emnlp.475
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2024
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
8122–8138
Language:
URL:
https://aclanthology.org/2024.findings-emnlp.475
DOI:
Bibkey:
Cite (ACL):
Linyan Yang, Jingwei Cheng, and Fu Zhang. 2024. Advancing Cross-Lingual Entity Alignment with Large Language Models: Tailored Sample Segmentation and Zero-Shot Prompts. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 8122–8138, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Advancing Cross-Lingual Entity Alignment with Large Language Models: Tailored Sample Segmentation and Zero-Shot Prompts (Yang et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-emnlp.475.pdf