Abstract
This research explores the intersection of formal knowledge representation and pure mathematics by investigating whether the Web Ontology Language (OWL), typically used for classifying web data and domain knowledge, can handle the rigorous, variable-heavy definitions of abstract algebra. The study attempts to model algebraic structures such as groups and rings using OWL 2 DL, identifying critical expressive limitations inherent in Description Logics when applied to universal mathematical properties.
Introduction
Ontologies are highly effective at categorizing hierarchical relationships—for example, defining that a "Penguin" is a subclass of "Bird." However, they are rarely employed to model the internal logic of mathematics. Abstract algebra is defined by universal properties: for a set to be classified as a "commutative ring," every element within that set must obey specific axioms, such as the commutative law ($a + b = b + a$). This paper investigates the challenges of translating these "for all" mathematical laws into the language of Description Logics (DL), with the ultimate goal of creating a system where a reasoner can automatically infer a structure's type and its applicable theorems.
Methodology
The ontology was developed using OWL 2 DL (specifically the $\mathcal{SHIQ}$ profile) within the Protégé 5.5.0 editor. To bridge the gap between algebraic definitions and DL, several modeling techniques were employed:
- Expression Classes: Individuals were created to represent specific operations and their results (e.g., an individual named
a*b). - Relational Constructs: The model leveraged transitive roles ($\mathcal{S}$), role hierarchies ($\mathcal{H}$), role inverses ($\mathcal{I}$), and qualified number restrictions ($\mathcal{Q}$) to define the relationships between elements through algebraic operations.
- Property Mapping: To simulate properties like commutativity, object properties (such as "commutes") were used to link expressions. However, this required extensive manual declaration to bypass the logic's inherent lack of variables.
As a theoretical modeling project, the study utilized standard algebraic definitions and prototype structures, such as the set of integers, rather than external empirical datasets.
Key Challenges in Mathematical Modeling
The Open World Assumption (OWA)
OWL operates under the Open World Assumption, which posits that if a statement is not explicitly known to be false, it may still be true. In the context of algebra, this creates a significant obstacle. A reasoner cannot definitively conclude that a structure is not commutative simply because it has not encountered a counterexample. To address this, the author proposed a "guilty until proven innocent" workaround, where a structure is assumed to be commutative unless an inconsistency is identified.
The Variable Gap
Description Logics are primarily designed for unary and binary predicates and lack the native variables required for complex equations like the associative law: $a * (b * c) = (a * b) * c$. Modeling these relationships required the creation of complex chains of individuals, a process that is both manually intensive and computationally expensive.
The Identity Element Problem
Defining an identity element (e.g., $0$ in addition) requires proving that $e * a = a$ for all elements $a$. Under OWA, a reasoner cannot be certain an element is truly the identity because there may be an "unseen" element for which the rule fails. Consequently, the identity must be manually declared by the user; the reasoner's role is limited to checking if that declaration remains logically consistent with the rest of the model.
Results and Discussion
The study achieved partial success in modeling the basic hierarchy of algebraic structures. The resulting ontology is capable of classifying declared structures and performing consistency checks. However, the project encountered a "scaling wall":
- Inference vs. Discovery: While OWL is excellent for classification, it struggles with discovery. It can verify if manual labels are logically sound but cannot autonomously "discover" an identity element or a latent property within a structure.
- Complexity of Associativity: While commutativity (involving two variables) was manageable through specific declarations, associativity (involving three variables) proved too complex to define effectively within the $\mathcal{SHIQ}$ framework.
Conclusion and Future Work
The research confirms that while OWL 2 DL can represent certain aspects of abstract algebra, significant concepts remain out of reach due to the constraints of standard Description Logics. Future work should address the "ubiquity of infinite sets" in mathematics, which current reasoners struggle to process. The author suggests that for ontologies to be truly useful in the mathematical domain, new modeling patterns or extensions to DL semantics—specifically those capable of handling "relative clause" type expressions—will be necessary to accurately capture algebraic definitions.
Summary updated