Tool Support for Ontology Engineering

Ian HorrocksFoundations for the Web of Information and Services 2011 · DOI

TL;DR

This paper highlights the critical role of sophisticated tool support in the development and maintenance of high-quality OWL ontologies, particularly for large-scale, real-world applications. It argues that while OWL's standardization has fostered a rich ecosystem of tools, these are not just useful but essential for managing the complexities of ontology engineering, enabling error detection, explanation, and efficient development.

What Problem Does This Paper Address?

The paper addresses the growing challenge of developing and maintaining complex ontologies, especially in domains like medicine and biology where ontologies are becoming indispensable for data management and integration. While the Web Ontology Language (OWL) has been standardized and offers formal semantics, building robust, error-free ontologies is a significant undertaking. This is particularly true for large ontologies developed by distributed teams over extended periods. The core problem is how to effectively support the engineering process to ensure the quality, consistency, and usability of these ontologies, moving beyond theoretical foundations to practical, large-scale application.

Methods

The paper doesn't introduce novel algorithms or methods in the traditional sense. Instead, it focuses on the application and integration of existing reasoning techniques and tools within the ontology engineering lifecycle. Key methods discussed include:

  • Description Logic (DL) Reasoning: The paper emphasizes how DLs provide the formal underpinning for OWL, enabling reasoning tasks like classification and query answering. It mentions two main types of reasoning algorithms: tableau algorithms (model-building) and resolution-based algorithms (proof-based).
  • Ontology Editors and Development Environments: Tools like Protégé-OWL, TopBraid Composer, and Neon are presented as crucial for the creation and manipulation of OWL ontologies.
  • Reasoning Systems: HermiT, FaCT++, Pellet, and RACER are highlighted as essential for checking ontology consistency, inferring new knowledge, and identifying errors.
  • Explanation and Justification Tools: These tools are vital for understanding why a reasoner reaches a particular conclusion or identifies an inconsistency, which is crucial for domain experts to trust and correct the ontology.
  • Ontology Mapping and Integration Tools: Tools like Prompt and ContentMap are mentioned for combining different ontologies.
  • Ontology Modularization and Extraction Tools: Tools like ProSÉ are discussed for creating smaller, manageable subsets of larger ontologies.
  • Ontology Version Control: Tools like ContentCVS are presented for managing changes to ontologies over time.

The paper's novelty lies in its comprehensive overview and advocacy for the synergistic use of these tools to overcome the practical challenges of ontology engineering.

Datasets

The paper does not introduce new datasets or benchmarks. Instead, it uses real-world ontologies as case studies and examples:

  • SNOMED CT: A large clinical ontology used extensively in healthcare systems.
  • GALEN: An open-source ontology for medical applications.
  • Foundational Model of Anatomy (FMA): An ontology of human anatomy.
  • National Cancer Institute (NCI) Thesaurus: An ontology for cancer diseases and treatments.
  • OBO Foundry: A repository of biomedical ontologies.

These are existing, large-scale ontologies that serve as the context for discussing the need for and impact of tool support.

Key Insights and Ideas

  • Tools are Essential, Not Optional: The paper strongly argues that for realistic, high-quality ontology development, sophisticated tool support is not merely beneficial but essential.
  • Reasoning is Key to Quality: The formal semantics of OWL, derived from description logics, enable powerful reasoning capabilities that are critical for detecting inconsistencies, under-constrained definitions, and modeling errors.
  • Explanation is Crucial for Trust: When reasoners identify errors or unexpected entailments, explanation tools are vital for domain experts to understand the cause, build trust in the system, and guide corrections. Without explanation, experts might dismiss the reasoning results.
  • Large-Scale Ontologies Magnify Challenges: The complexity and scale of ontologies like SNOMED CT necessitate advanced tools for modularization, integration, and version control to facilitate collaborative development and maintenance.
  • Beyond Academia: The paper emphasizes that the benefits of these tools are now recognized and utilized in commercial settings (e.g., British Telecom, Siemens), indicating their maturity and practical value.
  • Modeling Errors are Subtle: The paper illustrates how seemingly simple modeling choices (like defining junction regions as disjoint subclasses) can lead to complex inconsistencies when combined with other axioms, highlighting the need for careful, tool-assisted modeling.

Key Results

The paper doesn't present quantitative results in the form of experiments with specific metrics. Instead, its "results" are qualitative and based on observed benefits and case studies:

  • Improved Ontology Quality: Tool support, particularly reasoning, helps identify and correct errors, leading to more consistent and accurate ontologies. For example, the paper mentions that about 180 missing subclass relationships were detected in SNOMED CT using FaCT++.
  • Facilitation of Complex Modeling: The paper demonstrates how tools and reasoning support enable the expression of complex concepts (e.g., negative constraints, disjunctions, cardinality constraints) and the resolution of modeling issues that arise during extensions.
  • Enhanced Collaboration: Tools for modularization and version control support collaborative development efforts on large ontologies.
  • Increased Trust and Adoption: The availability of explanation facilities helps domain experts understand and trust the reasoning process, encouraging the adoption of OWL and related tools.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges the high worst-case complexity of OWL reasoning (NExpTime-complete for OWL and 2NExpTime-complete for OWL 2), although it notes that optimized algorithms perform well in practice.
  • It implicitly suggests that the "correct" modeling of complex anatomical regions (like "Groin") is challenging and requires expressive ontology languages and careful axiomatization.
  • The paper highlights that even with tools, identifying the source of errors and devising appropriate repairs can still be difficult, underscoring the need for effective explanation and repair mechanisms.
  • Future Work:
    • The paper implies a continued need for the development and refinement of tools that support the entire ontology engineering lifecycle, from creation to maintenance and integration.
  • It suggests that further research into explanation and repair mechanisms will be crucial for making ontology engineering more accessible and reliable.
  • The ongoing evolution of OWL (e.g., OWL 2) will likely necessitate further development of corresponding tool support.

Summary updated

BibTeX
@inproceedings{horrocks2011tool,
  author = {Ian Horrocks},
  title = {{Tool Support for Ontology Engineering}},
  booktitle = {Foundations for the Web of Information and Services},
  year = {2011},
  doi = {10.1007/978-3-642-19797-0\_6},
}
Figure