Meta-Semantics Augmented Few-Shot Relational Learning

TL;DR

This paper introduces PromptMeta, a meta-learning framework that leverages rich semantic information in knowledge graphs to improve few-shot relational learning, enabling better adaptation to new relations with limited data.

Contribution

The paper proposes a novel PromptMeta framework that integrates meta-semantics with relational info using a meta-semantic prompt pool and a dynamic fusion mechanism, advancing few-shot KG reasoning.

Findings

PromptMeta outperforms existing methods on real-world KG benchmarks.

Meta-semantics significantly enhance knowledge transfer in few-shot learning.

Dynamic fusion improves adaptation to new relations.

Abstract

Few-shot relational learning on knowledge graph (KGs) aims to perform reasoning over relations with only a few training examples. While current methods have focused primarily on leveraging specific relational information, rich semantics inherent in KGs have been largely overlooked. To bridge this gap, we propose PromptMeta, a novel prompted meta-learning framework that seamlessly integrates meta-semantics with relational information for few-shot relational learning. PromptMeta introduces two core innovations: (1) a Meta-Semantic Prompt (MSP) pool that learns and consolidates high-level meta-semantics shared across tasks, enabling effective knowledge transfer and adaptation to newly emerging relations; and (2) a learnable fusion mechanism that dynamically combines meta-semantics with task-specific relational information tailored to different few-shot tasks. Both components are optimized jointly with model parameters within a meta-learning framework. Extensive experiments and analyses on two real-world KG benchmarks validate the effectiveness of PromptMeta in adapting to new relations with limited supervision.