Scholarlink

With the rapid expansion of artificial intelligence applications, traditional entity alignment methods in knowledge graphs often face challenges in scalability and robustness, particularly when dealing with large and heterogeneous datasets. In this paper we propose EA-RAG, a novel Retrieval-Augmented Generation approach for entity alignment, combining retrieval mechanisms with large language models to explore generative solutions. Using the DBP15k dataset, we compare EA-RAG with several embedding-based models from the literature. Our results demonstrate significant potential for EA-RAG as a generative alternative to embedding-based methods in entity alignment tasks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Retrieval-Augmented Generation for Entity Alignment in Knowledge Graphs: An Incipient Experiment

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Davide Mario Ricardo Bara,
  • Daria Maria Mesesan,
  • Gheorghe Cosmin Silaghi

摘要

With the rapid expansion of artificial intelligence applications, traditional entity alignment methods in knowledge graphs often face challenges in scalability and robustness, particularly when dealing with large and heterogeneous datasets. In this paper we propose EA-RAG, a novel Retrieval-Augmented Generation approach for entity alignment, combining retrieval mechanisms with large language models to explore generative solutions. Using the DBP15k dataset, we compare EA-RAG with several embedding-based models from the literature. Our results demonstrate significant potential for EA-RAG as a generative alternative to embedding-based methods in entity alignment tasks.