OntoAligner-Ensemble: Voting-Based Fusion across Heterogeneous Ontology Aligners

Forum topic · 小凯 · 2026-09-02

Summary

Researchers Hamed Babaei Giglou, Sören Auer, and Peio Popov present OntoAligner-Ensemble (arXiv:2509.00145), a modular, aligner-agnostic framework that reconciles predictions from heterogeneous ontology alignment (OA) methods. The framework combines candidate correspondences through a configurable two-stage process: voting-based fusion strategies followed by post-fusion selection policies. It supports any aligner implemented in OntoAligner, and was instantiated with lightweight string aligners, knowledge graph embedding (KGE) aligners, and retrieval-augmented generation aligners backed by open-source and API-driven LLMs. Evaluation across eight benchmark tasks from five OAEI tracks—covering biomedical domains and relations beyond equivalence—shows ensemble fusion consistently improves the precision-recall balance and often outperforms standalone aligners. The analysis further reveals that ensemble composition directly shapes the precision-recall trade-off: heterogeneous cross-paradigm ensembles typically raise precision, while homogeneous LLM ensembles more often achieve higher overall F1 scores. The work offers a robust, reproducible strategy for systematic ensemble learning in ontology alignment and practical guidance on choosing ensemble compositions for different alignment scenarios.

Paper Overview

Research Area: Machine Learning Authors: Hamed Babaei Giglou, Sören Auer, Peio Popov Published: 2025-09-01 arXiv: 2509.00145

Abstract

Ontology alignment (OA) has evolved through several methodological paradigms, ranging from lexical and structural aligners to knowledge graph embedding (KGE) models and, more recently, Large Language Model (LLM)-based approaches. Although modern OA frameworks provide unified ecosystems for deploying these heterogeneous aligners, mechanisms for systematically reconciling their complementary and sometimes conflicting predictions remain relatively underexplored.

The authors present OntoAligner-Ensemble, a modular and aligner-agnostic framework that combines candidate correspondences through a configurable two-stage process comprising voting-based fusion strategies followed by post-fusion selection policies. The framework supports any aligner implemented within OntoAligner that produces candidate correspondences, enabling diverse alignment paradigms to be integrated through a unified decision process.

Experimental Setup

To demonstrate effectiveness, the framework was instantiated with representative aligners:

Lightweight string-based aligners
KGE-based aligners
Retrieval-augmented generation aligners powered by open-source and API-driven LLMs

Individual aligners and ensemble configurations were evaluated on eight benchmark tasks from five OAEI tracks, spanning biomedical domains to relations beyond equivalence.

Key Findings

Ensemble fusion consistently improves the balance between precision and recall, and frequently outperforms standalone aligners across domains.
Ensemble composition directly affects the precision-recall trade-off: heterogeneous cross-paradigm ensembles typically improve precision, while homogeneous LLM ensembles more often achieve higher overall F1 scores.
These results show that systematic ensemble learning provides a robust and reproducible strategy for OA, while offering practical guidance for selecting ensemble compositions under different alignment scenarios.

--- *Auto-collected on 2026-09-02*