In traditional databases, a schema acts as a gatekeeper. If you try to insert a "Person" record without a "Birthdate," the database rejects it. This is the Closed-World Assumption (CWA): if a fact is not present in the database, it is assumed to be false or invalid.
However, the Web Ontology Language (OWL) and the Semantic Web operate on the Open-World Assumption (OWA). In this view, if a birthdate is missing, the system doesn't assume the person was never born; it simply assumes the information is currently unknown. As Boris Motik, Ian Horrocks, and Ulrike Sattler explain in Bridging the gap between OWL and relational databases, this creates a fundamental "semantic friction." OWL axioms are deductive rules that fill in gaps, while database schemas are integrity constraints that flag errors.
The Problem of Invented Facts
The OWA is powerful for linking data across the web, but it is disastrous for data validation. If you define a "Valid Entry" as something that must have a timestamp, an OWL reasoner won't flag an entry missing a timestamp as an error. Instead, it will logically infer that a timestamp must exist somewhere, even if it isn't in your dataset.
This mismatch is a major hurdle in fields like biology and manufacturing. Alan Ruttenberg and colleagues noted in What BioPAX Communicates and how to extend OWL to help it that biological databases need immediate error reporting when mandatory records are omitted. Without a way to "close" the world, the logic system simply invents facts to satisfy its own rules, rendering it useless for quality control.
Bridging the Gap with Epistemic Logic
To solve this, researchers have moved toward "hybrid" semantics that allow OWA and CWA to coexist. One influential approach involves epistemic logic—shifting the focus from what is "true in the world" to what the system "actually knows."
Marco Console and Maurizio Lenzerini argue in Epistemic Integrity Constraints for Ontology-Based Data Management that constraints should be checked against the system's current state of knowledge. If the system cannot prove a timestamp exists from the available data, the constraint is violated. This prevents the reasoner from "hallucinating" missing data to satisfy a schema.
Similarly, Jiao Tao and Deborah L. McGuinness proposed a new Integrity Constraint semantics for OWL that adopts a "weak" Unique Name Assumption. This allows developers to perform rigorous validation while still benefiting from standard open-world reasoning for other tasks.
Practical Implementation: DBoxes and Rewriting
How do we actually run these checks on large-scale data? Several researchers suggest "closing" only specific parts of the ontology.
- DBoxes: Inanç Seylan and Enrico Franconi introduced the concept of DBoxes, where certain predicates are treated as "complete" database tables. This allows complex ontological queries to be rewritten into standard SQL.
- Combined Approaches: Carsten Lutz and Frank Wolter developed techniques to rewrite both data and queries into formats compatible with relational databases like IBM DB2, bridging the logical gap without losing the ability to make inferences.
- Constraint Rewriting: In Validation of SHACL Constraints over KGs with OWL 2 QL Ontologies via Rewriting, Evgeny Kharlamov and Ognjen Savković show how to validate modern constraint languages (like SHACL) by accounting for facts that are logically implied by an ontology but not explicitly written in the database.
Why This Matters
The goal of this research is a "best of both worlds" scenario. Frameworks like the Linked Data Modeling Language (LinkML), described by Sierra A. T. Moxon and James P. Balhoff, embody this direction. By grounding developer-friendly formats in machine-readable semantics, researchers can ensure data integrity (CWA) while maintaining the global interoperability and flexibility of the Semantic Web (OWA).
As industrial applications grow, tools like SOMM are putting these theories into practice, allowing engineers to enforce the "hard" constraints required for manufacturing while still using the "soft" reasoning of ontologies to discover new insights.