Fusing Context and Path with Efficient Negative Sampling and Retrieval for Inductive Link Prediction

Xinyu Liang

doi:10.1016/j.eswa.2025.129894

Fusing Context and Path with Efficient Negative Sampling and Retrieval for Inductive Link Prediction

2025-10-20

0

PRE

AI

DOI：10.1016/j.eswa.2025.129894

Original

Original request for help

Save

Abstract

En 中文

• Fusing Context and Path with efficient Negative sampling and Retrieval (CPNR). • Variants CPNR R and CPNR S use random and semantic-aware negative sampling. • Paths and contexts are merged with a novel context extraction and filtering module. • Tailored prompts for large language models extend short entity descriptions. • CPNR S and CPNR R excel in transductive, inductive, and few-shot settings.

Fusing Context and Path with Efficient Negative Sampling and Retrieval for Inductive Link Prediction

Abstract

Journal

Organization