Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Complex Ontology Alignment using Large Language Models

Reihaneh Amini, Sanaz Saki Norouzi, Pascal Hitzler, Reza Amini

TL;DR

The paper tackles the challenge of automating complex ontology alignment beyond simple 1-to-1 mappings. It introduces a prompt-based LLM approach that leverages rich module content from the GeoLink Modular Ontology (GMO) and its interactions with the GeoLink Base Ontology (GBO). Evaluation on the GeoLink Complex Alignment dataset with GPT-4 shows substantial gains when module information is included, with a mean recall/precision around $0.67$ and a median near $0.75$–$0.80$, and a substantial portion achieving perfect scores. The study demonstrates a neural-symbolic pathway for complex alignment that does not rely on shared individuals and highlights the critical role of ontology structure in enabling automated reasoning. Future work will pursue human-in-the-loop workflows, richer module representations, and integration with symbolic reasoning and traditional alignment methods to broaden robustness and autonomy.

Abstract

Ontology alignment, a critical process in the Semantic Web for detecting relationships between different ontologies, has traditionally focused on identifying so-called "simple" 1-to-1 relationships through class labels and properties comparison. The more practically useful exploration of more complex alignments remains a hard problem to automate, and as such is largely underexplored, i.e. in application practice it is usually done manually by ontology and domain experts. Recently, the surge in Natural Language Processing (NLP) capabilities, driven by advancements in Large Language Models (LLMs), presents new opportunities for enhancing ontology engineering practices, including ontology alignment tasks. This paper investigates the application of LLM technologies to tackle the complex ontology alignment challenge. Leveraging a prompt-based approach and integrating rich ontology content so-called modules our work constitutes a significant advance towards automating the complex alignment task.

Towards Complex Ontology Alignment using Large Language Models

TL;DR

The paper tackles the challenge of automating complex ontology alignment beyond simple 1-to-1 mappings. It introduces a prompt-based LLM approach that leverages rich module content from the GeoLink Modular Ontology (GMO) and its interactions with the GeoLink Base Ontology (GBO). Evaluation on the GeoLink Complex Alignment dataset with GPT-4 shows substantial gains when module information is included, with a mean recall/precision around and a median near –, and a substantial portion achieving perfect scores. The study demonstrates a neural-symbolic pathway for complex alignment that does not rely on shared individuals and highlights the critical role of ontology structure in enabling automated reasoning. Future work will pursue human-in-the-loop workflows, richer module representations, and integration with symbolic reasoning and traditional alignment methods to broaden robustness and autonomy.

Abstract

Ontology alignment, a critical process in the Semantic Web for detecting relationships between different ontologies, has traditionally focused on identifying so-called "simple" 1-to-1 relationships through class labels and properties comparison. The more practically useful exploration of more complex alignments remains a hard problem to automate, and as such is largely underexplored, i.e. in application practice it is usually done manually by ontology and domain experts. Recently, the surge in Natural Language Processing (NLP) capabilities, driven by advancements in Large Language Models (LLMs), presents new opportunities for enhancing ontology engineering practices, including ontology alignment tasks. This paper investigates the application of LLM technologies to tackle the complex ontology alignment challenge. Leveraging a prompt-based approach and integrating rich ontology content so-called modules our work constitutes a significant advance towards automating the complex alignment task.
Paper Structure (11 sections, 2 equations, 7 figures, 2 tables)

This paper contains 11 sections, 2 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Complex Alignment by Large Language Model
  4. Ontology Modules
  5. Design of the Prompting Process
  6. Evaluation
  7. Prompting and Detailed Example
  8. Quantitative Evaluation
  9. Discussion and Future Work
  10. Conclusion
  11. Acknowledgments

Figures (7)

  • Figure 1: A diagram for a Person module
  • Figure 2: Prompting workflow
  • Figure 3: Uploading GMO.ttl file as prompt.
  • Figure 4: GBO-related instruction and question in prompt
  • Figure 5: GPT-4 response to our initial question
  • ...and 2 more figures