Graph Fusion Across Languages using Large Language Models

TL;DR

This paper introduces a novel framework that uses Large Language Models to fuse multiple knowledge graphs across languages by translating graph triplets into natural language for better semantic alignment.

Contribution

The paper presents a new cross-lingual graph fusion method leveraging LLMs' reasoning and multilingual priors, enabling scalable and modular knowledge integration across languages.

Findings

LLMs can serve as a universal semantic bridge for cross-lingual graph fusion.

The framework successfully merges heterogeneous graphs in a sequential, scalable manner.

Experimental results on DBP15K demonstrate effective cross-lingual knowledge synthesis.

Abstract

Combining multiple knowledge graphs (KGs) across linguistic boundaries is a persistent challenge due to semantic heterogeneity and the complexity of graph environments. We propose a framework for cross-lingual graph fusion, leveraging the in-context reasoning and multilingual semantic priors of Large Language Models (LLMs). The framework implements structural linearization by mapping triplets directly into natural language sequences (e.g., [head] [relation] [tail]), enabling the LLM to map relations and reconcile entities between an evolving fused graph ($G_{c}^{(t-1)}$) and a new candidate graph ($G_{t}$). Evaluated on the DBP15K dataset, this exploratory study demonstrates that LLMs can serve as a universal semantic bridge to resolve cross-lingual discrepancies. Results show the successful sequential agglomeration of multiple heterogeneous graphs, offering a scalable, modular solution for continuous knowledge synthesis in multi-source, multilingual environments.