A Modular Consistency Proof for DOLCE

Oliver Kutz, Till MossakowskiAAAI 2011 · DOI

Introduction

In formal artificial intelligence, foundational ontologies like the Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE) provide the essential conceptual framework for machine understanding of time, parthood, and physical objects. However, ensuring the logical consistency of such massive systems—often comprising hundreds of first-order logic axioms—is a significant challenge. Standard automated reasoning tools and model finders frequently fail to scale, making global consistency proofs notoriously difficult to achieve. This paper presents a novel modular strategy that successfully establishes a global consistency proof for DOLCE by decomposing it into manageable sub-models.

The Problem of Scale in Ontology Verification

The primary obstacle in verifying large-scale ontologies is the exponential complexity of finding a valid mathematical structure (a model) where all axioms hold true. Inconsistencies can be introduced easily during the design phase, yet standard model finders may churn for weeks on even a small subset of DOLCE without reaching a conclusion. Manual proofs are similarly impractical due to the intricate web of interdependencies between concepts.

Methodology: Architectural Specifications and Conservativity

To address these scaling issues, the authors adopt a "divide and conquer" approach inspired by architectural specification techniques in software engineering. The methodology involves three critical phases:

  1. Decomposition: The DOLCE ontology was manually decomposed into 38 distinct sub-theories or units.
  2. Relative Consistency via Conservativity Triangles: Instead of proving the consistency of the entire system at once, the authors used "conservativity triangles." By demonstrating that a complex theory $T$ is a conservative extension of a simpler, already-proven theory $T'$, the consistency of $T$ is established relative to $T'$.
  3. Amalgamation and Verification: The Heterogeneous Tool Set (HETS) was employed to verify "amalgamability," ensuring that when sub-models are merged, their shared symbols do not create logical conflicts.
A conservativity triangle illustrating the proof of relative consistency between theories.

A significant technical shift in this work was the move from DOLCE’s traditional top-down design to a bottom-up construction. This required restructuring the mereology (the logic of parts and wholes) to allow for the aggregation of "leaf" concepts into higher-level ontological structures.

Key Technical Innovations

The proof process necessitated several innovative workarounds to handle complex logical dependencies:

  • The Abstraction Barrier: By enforcing an abstraction barrier, the system only interacts with the properties of sub-units rather than their internal logic, preventing "spaghetti dependencies" that would otherwise break the proof.
  • The "Eternal Object" Hack: To facilitate the verification of temporal properties, the authors introduced a hypothetical "eternal object" that exists across all time points. This simplification allowed automated reasoners like SPASS to successfully verify local consistency.

Results and Impact

The authors successfully established the first comprehensive modular consistency proof for the entirety of DOLCE. The effort was quantified by the decomposition into 38 units, with 18 amalgamations automatically verified by HETS. Only one unit (Time Mereology) was solved directly by a model finder, highlighting the necessity of the modular approach for the remaining 37 units.

The dependency graph showing the bottom-up construction of the DOLCE ontology from modular sub-theories.

This work demonstrates that tools and methodologies from software verification, such as CASL and HETS, are highly effective for solving complex problems in formal logic and ontology.

Limitations and Future Directions

While the verification phase is automated, the initial architectural specification—the blueprint for decomposition—remains a high-effort manual task requiring expert knowledge. Furthermore, the rigorous proof process uncovered potential "design flaws" in DOLCE, where overly restrictive sub-concepts may limit future refinements. Future research aims to scale these modular techniques to even larger ontologies, such as the Suggested Upper Merged Ontology (SUMO), which remains beyond the reach of current monolithic verification methods.

Summary updated

BibTeX
@inproceedings{kutz2011modular,
  author = {Oliver Kutz and Till Mossakowski},
  title = {{A Modular Consistency Proof for DOLCE}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2011},
  doi = {10.1609/aaai.v25i1.7860},
}
Figure