Why shared meaning matters

A date is not yet an explanation

Imagine receiving “survey date: 20 August 2026” for Harbour Court. The date is readable and correctly formatted. Yet you cannot tell whether someone visited the dwelling that day, issued a report, corrected an earlier report or entered information into a system. The value may pass a format check while answering a different question from the one you intended to ask.

In our fictional case, an inspection of Flat 1 occurred on 12 August. Report version 1 was issued on 14 August. Version 2, issued on 20 August, corrected the description of an access limitation; no second visit occurred. Calling all three dates “survey date” discards distinctions before information is exchanged. The problem is not the date format. It is the missing explanation of what each date describes.

Semantic modelling is the work of making those meanings explicit: what we are describing, what we say about it, how it relates to other things and under which conditions a statement applies. An ontology formally expresses the selected distinctions. It does not replace professional knowledge; it gives that knowledge a form people can examine, challenge and use consistently.

The world, the agreement and a particular account

Keep three levels apart. There is a dwelling in the world. There is a model definition explaining what the model means by a dwelling. And there are particular records describing it. A definition can be clear while a record contains an error. A record can change while the dwelling remains the same. Neither a model nor a collection of records is the world itself.

Harbour Court is a building containing Flat 1 and Flat 2. An address description helps people locate or refer to a dwelling. A title record serves another purpose. A report describes an inspection and its findings. The appearance of “14 Orchard Road” in each description does not make these subjects interchangeable. Correcting an address's spelling does not create a dwelling; keeping its spelling does not prove that a later subdivision left the same dwelling in existence.

These are questions about resources—the identifiable things and concepts we describe—and their identity. A stable reference helps us talk about a subject repeatedly. It cannot decide, on its own, whether two descriptions concern the same subject.

Read the connection as a sentence

Start without any diagram: Report version 2 describes the 12 August inspection. The first part tells us which subject we are talking about. “Describes” names the relationship. The last part tells us what that subject connects to. The diagram is another way of showing that sentence.

Direction matters. Read from the arrow's starting point to its end. Reversing it would say that the inspection describes the report, which is not the claim intended here. The relationship name matters too: replacing “describes” with “is the same as” would make a different and unjustified claim. An unlabelled line leaves the reader to guess.

Some sentences need more context than two names and a verb. “Alex is a buyer” leaves the arrangement unstated; “Alex participates as buyer in the proposed sale of Flat 1” identifies it. A relationship can also need a period, source or purpose. The first review task is to notice the missing meaning, not to choose how a technical specialist should encode it.

One account, three useful representations

A form can be convenient for entering or reading a report. A nested description groups information beneath a chosen starting subject. A graph draws out connections across subjects. Each can express the same facts when its fields, references and relationships have clear meanings. Here is one deliberately complete account, shown three ways.

The graph helps someone follow a subject across descriptions. It does not guarantee correct claims, automatically eliminate duplicates or prove that two similarly labelled nodes are identical. Equally, a familiar form is not inherently less meaningful. The test is what a recipient can understand after the information has moved between these representations.

The benefit comes from preserving a useful distinction

Fewer repeated explanations, traceable evidence and understandable corrections are worthwhile aims. They depend on maintained definitions and exchanges that preserve the agreement. The benefits, limits and reasonable objections chapter examines these conditions, costs and alternatives in full.

Agreeing meaning is not centralising everybody's data

People can share definitions and identifiers without placing every property record in one database. A model does not prove that an inspection happened, grant access to a report or make an AI draft an approved standard. Accuracy, access and adoption need their own evidence. Your useful contribution is a precise question about what the agreement says.

If you make formal modelling decisions, the separate classes and relationships chapter explains the technical representation. You do not need its syntax to review the sentences on this page.