Toward Multi-viewpoint Reasoning with OWL Ontologies

Heiner StuckenschmidtLecture Notes in Computer Science 2006 · DOI · OpenAlex

Abstract

Ontologies are often designed to be task-neutral, yet practical applications frequently require task-specific perspectives. This paper introduces a formal framework for "multi-viewpoint reasoning" within the $\mathcal{SIN}$ description logic. By defining a vocabulary-limited approximation that replaces irrelevant concepts and roles with the universal concept ($\top$), the approach enables the extraction of simplified hierarchies tailored to specific application needs. This method operates within standard OWL-DL semantics, avoiding the need for language extensions while providing a computationally efficient way to manage ontological complexity.

Introduction

The promise of ontologies lies in their ability to provide a shared, reusable description of a domain. However, this "task-neutrality" often conflicts with the specific requirements of individual applications. A comprehensive ontology may contain excessive detail, leading to unnecessary computational overhead or "unwanted conclusions" that are irrelevant or even legally problematic for a given task.

For instance, a public sector salary system may need to track parental status for benefits but must remain gender-blind, whereas a pension system may require gender data for retirement calculations. Current reasoning engines force both systems to process the full complexity of the underlying model. This paper proposes a mechanism to "switch off" parts of the vocabulary to generate application-specific viewpoints.

Methodology

The proposed framework centers on the concept of vocabulary-limited approximation. Given a sub-vocabulary $V'$, any concept name or role not present in $V'$ is replaced by $\top$ (the "top" concept). This process effectively "blurs" the distinctions made by the excluded terms.

A new operator, subsumption with respect to a sub-vocabulary ($\sqsubseteq_{V'}$), is introduced. A concept $C$ is subsumed by $D$ relative to $V'$ if the logical implication holds after the irrelevant vocabulary has been stripped. To implement this efficiently, the paper defines Maximal Subsumption Vocabularies (MSVs)—the largest subsets of a vocabulary that allow a specific subsumption relationship to persist. By pre-calculating these MSVs, the system can extract specific viewpoints in linear time relative to the number of concept pairs.

The Family Ontology Example

To illustrate these concepts, the author utilizes a constructed "Family Ontology" containing concepts such as Person, Man, Woman, Mother, and Grandmother, along with roles like has-gender and has-child.

The full hierarchy of the Family Ontology example, showing distinct gendered and relational branches.

In its full form, the ontology maintains distinct branches for gendered and relational concepts. However, when an application requires a gender-neutral viewpoint, the proposed method allows for a systematic simplification of this structure.

Key Insights and Results

The primary effect of limiting vocabulary is hierarchical collapse. When gender-related terms are ignored, concepts like Man and Woman become logically indistinguishable from Person. This is not a failure of the logic, but a deliberate feature reflecting the application's indifference to those specific distinctions.

The collapsed hierarchy resulting from ignoring gender-specific vocabulary, where gendered roles are grouped together.

As shown in the figures, the complex initial hierarchy simplifies into a grouped structure where Mother, Father, and Parent are treated as equivalent. Crucially, this approach requires no extensions to the OWL-DL language; it is a method of utilizing existing logic differently. Furthermore, the monotonicity of viewpoints—where a subsumption holding for a large vocabulary must hold for its subsets—allows for significant optimization in reasoning tasks.

Limitations and Future Work

While promising, the current model has limitations regarding language coverage. It excludes Role Hierarchies, Qualified Number Restrictions, and Nominals, as ignoring vocabulary in these contexts can inadvertently make expressions more specific, breaking the approximation model. Additionally, the initial computation of MSVs may be computationally intensive for very large ontologies. Future research will focus on a model-theoretic characterization to incorporate the currently excluded OWL-DL features.

Conclusion

This research provides a robust theoretical foundation for multi-viewpoint reasoning. By allowing applications to define their own relevant sub-vocabularies, the framework bridges the gap between general-purpose ontologies and task-specific requirements, enhancing the practical utility and reuse of complex knowledge bases.

Summary updated

BibTeX
@inproceedings{stuckenschmidt2006toward,
  author = {Heiner Stuckenschmidt},
  title = {{Toward Multi-viewpoint Reasoning with OWL Ontologies}},
  booktitle = {Lecture Notes in Computer Science},
  year = {2006},
  doi = {10.1007/11762256\_21},
}
Figure