LogMap family participation in the OAEI 2018

Ernesto Jiménez-Ruiz, Bernardo Cuenca Grau, Valerie CrossOM@ISWC 2018 · OpenReview

Introduction

LogMap is a veteran, logic-based framework for ontology matching designed to scale to massive knowledge bases while maintaining strict logical consistency. In the 2018 Ontology Alignment Evaluation Initiative (OAEI) campaign, the LogMap family demonstrated its versatility by introducing automated task division and integration with the HOBBIT benchmarking platform. It remains one of the few systems capable of identifying and "repairing" logical contradictions in real-time during the alignment process.

The Challenges of Modern Ontology Matching

In the Semantic Web ecosystem, different organizations often develop independent ontologies—structured frameworks of concepts and relationships. To enable interoperability between these systems, we must perform ontology matching: the process of finding correspondences (mappings) between equivalent concepts across different datasets.

This process faces two primary hurdles:

  1. Computational Scale: Modern ontologies can contain hundreds of thousands of classes. A naive comparison of every class to every other class results in a combinatorial explosion that is computationally unfeasible.
  2. Logical Consistency: Simple lexical matches (e.g., matching "Person" in one ontology to "Human" in another) can lead to "unsatisfiable classes." These are logical contradictions that break the integrity of the combined knowledge base.

While many matching systems focus solely on finding the most likely correspondences, they often ignore the logical fallout. LogMap addresses both challenges by combining high-speed lexical indexing with rigorous logic-based repair.

Methodology and System Architecture

LogMap is not a single algorithm but a suite of integrated techniques designed for efficiency and soundness:

  • Lexical Indexation: The system utilizes an inverted index—similar to search engine technology—to rapidly identify potential matches based on labels, which drastically reduces the initial search space.
  • Logic-based Module Extraction: Rather than reasoning over a massive ontology in its entirety, LogMap extracts smaller "modules" containing only the relevant semantic context for the classes being matched.
  • Propositional Horn Reasoning: The matching problem is encoded into Horn logic and processed using the Dowling-Gallier algorithm. This allows the system to detect unsatisfiable classes both soundly and efficiently.
  • Greedy Local Repair: When a contradiction is detected, LogMap performs an on-the-fly repair by discarding the problematic mapping, ensuring the resulting knowledge graph remains coherent.
  • Ontology Division: New for the 2018 campaign, this module uses neural embeddings and lexical indices to split massive matching tasks into smaller, independent subtasks that can be processed in parallel or more easily managed.

OAEI 2018 Participation and Results

The LogMap family—including the lightweight LogMapLt and the BioPortal-integrated LogMapBio—was evaluated against the OAEI 2018 tracks. These tracks typically include benchmarks ranging from conference data to massive biomedical ontologies such as SNOMED CT and the Foundational Model of Anatomy (FMA).

Key Insights

  • Logic as a Priority: LogMap treats logical coherence as a requirement rather than an afterthought, ensuring that merged ontologies are actually functional for reasoning.
  • The Repair Service: The authors have decoupled the repair module (LogMap-Repair) so it can be used as a standalone component by other matching systems that lack native reasoning capabilities.
  • Divide and Conquer: The introduction of the Ontology Division module suggests that the future of large-scale matching lies in intelligent task decomposition using machine learning to guide the split.
  • Accessibility: Beyond the competition, the authors maintain a web interface that has processed over 3,000 requests, highlighting the practical demand for user-interactive matching tools.

Performance Highlights

LogMap participated in almost all OAEI 2018 tracks, proving its robustness across diverse domains. It remains a top-tier performer, particularly in tracks where "coherent" (logically sound) mappings are required. Additionally, the system was successfully adapted for the HOBBIT platform, a specialized environment for benchmarking the performance of Big Linked Data systems.

Limitations and Future Work

The primary limitation of the current framework is its Lexical Dependency. LogMap relies heavily on the similarity between the vocabularies used in the input ontologies. If two ontologies are "lexically disparate"—describing the same concepts using entirely different terminology or lacking descriptive labels—LogMap is at a disadvantage. Future work will likely focus on bridging this lexical gap through deeper semantic analysis and the integration of external background knowledge.

Summary updated

BibTeX
@inproceedings{jimenezruiz2018logmap,
  author = {Ernesto Jiménez-Ruiz and Bernardo Cuenca Grau and Valerie Cross},
  title = {{LogMap family participation in the OAEI 2018}},
  booktitle = {OM@ISWC},
  year = {2018},
  url = {https://openreview.net/forum?id=v4pVNJA04Z},
}
Figure