OCOM5201M / Knowledge Representation and Reasoning

The Sandwich Ontology

My ontology assessment became a surprisingly serious investigation into sandwichness: bread, fillings, open-world assumptions, disjoint classes, cardinality constraints, doorstoppers, open sandwiches and the family debate over starch-on-starch-on-starch.

It was funny on the surface, but technically useful underneath: OWL, description logic, reasoners, consistency, modelling trade-offs, Lean and Prolog all made the same point in different dialects. Formal meaning is hard.

Screenshot from sandwich ontology assessment showing an open sandwich counterexample.
A sandwich ontology: ridiculous enough to remember, strict enough to hurt

Assessment

Define “sandwich”.

I modelled sandwiches, layers, bread, fillings and edge cases in OWL, then used a reasoner to classify valid and invalid instances.

Open world

Do not mistake absence for falsehood

The open-world assumption was the most practically unsettling idea: unknown facts are not automatically false, which changes how constraints must be expressed.

Formal tools

Lean and Prolog

Lean and Prolog exercises made inference rules tangible, while Protégé made modelling errors immediately visible.

Lesson

Meaning needs engineering

Ontologies are not mere taxonomy. They are executable claims about a domain, and bad modelling gives you bad reasoning with great confidence.