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

Forum topic · 小凯 · 2026-09-02

Summary

OntoAligner-Ensemble is a modular, aligner-agnostic framework for ontology alignment (OA) that systematically combines predictions from heterogeneous aligners—lexical/string-based, knowledge graph embedding (KGE), and retrieval-augmented LLM-based approaches—via a configurable two-stage process of voting-based fusion followed by post-fusion selection policies. Developed by Hamed Babaei Giglou, Sören Auer, and Peio Popov (arXiv:2509.00145, September 2025), the framework supports any aligner implemented in OntoAligner that produces candidate correspondences. Evaluations on eight benchmark tasks from five OAEI tracks, spanning biomedical to non-equivalence relations, show that ensemble fusion consistently improves the precision-recall balance and often outperforms individual aligners across domains. The analysis reveals that ensemble composition directly shapes the precision-recall trade-off: heterogeneous cross-paradigm ensembles typically boost precision, while homogeneous LLM ensembles more often achieve higher overall F1 scores. The work provides robust, reproducible strategies for systematic ensemble learning in OA and practical guidance on selecting ensemble compositions for different alignment scenarios.

Overview

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

Abstract (translated)

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: 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.

To demonstrate effectiveness, the framework was instantiated with representative lightweight string aligners, KGE-based aligners, and retrieval-augmented generation aligners powered by both open-source and API-driven LLMs. Individual aligners and ensemble configurations were evaluated on eight benchmark tasks from five OAEI tracks, covering domains from biomedical ontologies to relations beyond equivalence.

Key findings

Ensemble fusion consistently improves the balance between precision and recall.
Ensembles frequently outperform individual aligners across different domains.
Ensemble composition directly affects the precision-recall trade-off:
Heterogeneous cross-paradigm ensembles generally improve precision.
Homogeneous LLM ensembles more often achieve higher overall F1 scores.

Conclusion

The results indicate that systematic ensemble learning offers a robust and reproducible strategy for ontology alignment, while providing practical guidance for choosing ensemble compositions in different alignment scenarios.

---

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