Few-shot Link Prediction on N-ary Facts

TL;DR

This work defines Few-shot Link Prediction on Hyper-relational Facts (FSLPHFs) and presents MetaRH, a meta-learning model with three modules—Relation Learning, Support-specific Adjustment, and Query Inference—that derives meta relational information from limited support data and background facts. It introduces three datasets (F-WikiPeople, F-JF17K, F-WD50K) tailored for FSLPHFs and demonstrates that MetaRH outperforms representative LPHFs and FSLPBFs across these datasets, with competitive performance against a Large Language Model baseline in domain-specific settings. The results underscore the value of gradient-guided meta-relational adjustments and background-informed representations for robust reasoning under few-shot constraints, enabling more effective enrichment of hyper-relational knowledge graphs in dynamic environments. The work points toward integrating LLMs to further reduce dependence on extensive background data and task-specific training while maintaining strong reasoning capabilities in hyper-relational contexts.

Abstract

Hyper-relational facts, which consist of a primary triple (head entity, relation, tail entity) and auxiliary attribute-value pairs, are widely present in real-world Knowledge Graphs (KGs). Link Prediction on Hyper-relational Facts (LPHFs) is to predict a missing element in a hyper-relational fact, which helps populate and enrich KGs. However, existing LPHFs studies usually require an amount of high-quality data. They overlook few-shot relations, which have limited instances, yet are common in real-world scenarios. Thus, we introduce a new task, Few-Shot Link Prediction on Hyper-relational Facts (FSLPHFs). It aims to predict a missing entity in a hyper-relational fact with limited support instances. To tackle FSLPHFs, we propose MetaRH, a model that learns Meta Relational information in Hyper-relational facts. MetaRH comprises three modules: relation learning, support-specific adjustment, and query inference. By capturing meta relational information from limited support instances, MetaRH can accurately predict the missing entity in a query. As there is no existing dataset available for this new task, we construct three datasets to validate the effectiveness of MetaRH. Experimental results on these datasets demonstrate that MetaRH significantly outperforms existing representative models.

Figures (5)

• Figure 1: The histogram of relation frequencies in WD50K.
• Figure 2: The overview of MetaRH model. To distinguish the different elements in hyper-relational facts, we use red for relations, blue for head and tail entities, pink for attributes, and purple for values.
• Figure 3: An example of ChatGPT prompts for FSLPHFs.
• Figure 4: Impact of the few-shot size on the F-WD50K dataset.
• Figure 5: Impact of the number of background facts per entity and relation weight on the F-WD50K dataset.

Theorems & Definitions (3)

• Definition 1
• Definition 2
• Definition 3