LinkedIn ontology engineering synthesis

Ontology Is Source Code

Tony Seale argues that an ontology is not a database table or passive schema. It is formal logic and source code for organizational meaning, so it should compile, live in plain text, and be governed like code.

338Visible reactions at capture time.
37Visible comment count at capture time.
25Visible repost count at capture time.

Core Claims

The generated graph models the post as an engineering governance argument for ontology source management.

It should compile

Syntax, satisfiability, expected entailments, contradictions, and duplicate concepts belong in build checks.

Ontology Engineering Workflow

Use the same engineering control plane that already handles code carrying less conceptual weight.

Compile gate

A validation gate where ontology syntax, satisfiability, entailments, contradiction checks, and duplicate concept checks must pass.

Automated ontology checks

Build checks that catch syntax errors, unsatisfiable classes, unexpected entailments, contradictions, and duplicate concepts.

Governed like code

Ontology governance using branches, pull requests, reviews, automated checks, releases, and builds.

LLM-readable source

A design principle where large language models can read, navigate, and edit ontology source directly like a codebase.

Discussion Layer

Visible LinkedIn comments are represented as RDF comments and topic links, starting with Kingsley Uyi Idehen.

Kingsley Uyi Idehen11m · 0 reactions · 8 replies

Yes! The creation and management of ontologies, and even instance data, is a classic files-and-filesystem interaction pattern. Before LLMs this was difficult because RDF notations had tradeoffs: Turtle was human-readable but lacked tooling support, JSON-LD had more tooling support but was less readable, and RDF/XML was challenged on both fronts. Today, the Semantic Web stack gels naturally with LLMs and AI Agents. An agent can express a conceptualization in the notation you prefer and save it to a file; on suitable platforms those files can be copied to a folder transparently bound to a backend RDF DBMS for updating the triple or quad store.

Files-and-filesystem pattern RDF notation tradeoffs LLM-powered agents Folder-to-quad-store binding
Nicolas Matentzoglu2h · 0 reactions · 1 replies

For huge, highly regular ontologies such as anatomy, biological traits, or chemistry, one can sometimes curate the ontology in a table and map every instance or class to a design pattern that provides the logical modelling.

Table-driven ontology pattern Ontology

Knowledge Graph Explorer

Graph data is derived from the generated RDF entity and relationship model. Node and edge clicks use URIBurner resolver-backed IRIs.

RDF Graph Workbench

Explore ontology-as-code entities, practices, query examples, provenance, and source concepts.

0 nodes / 0 links

Graph data embedded from companion RDF at generation time. The controls tray is closed by default; Advanced exposes settings and predicate filters.

Executable SPARQL

Generated live queries use the companion named graph IRI.

Core Practices Run live query 
PREFIX schema: <http://schema.org/>
PREFIX ex: <https://www.linkedin.com/posts/tonyseale_did-you-start-building-your-ontology-as-a-share-7463353332565356545-jY54#>
SELECT ?practice ?name ?description
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontology-is-source-code-tony-seale-gpt5-chat-1.ttl> {
    ex:ontologyAsCode ex:hasComponent ?practice .
    ?practice schema:name ?name ;
              schema:description ?description .
  }
}
ORDER BY ?name
Compile Gates Run live query 
PREFIX schema: <http://schema.org/>
PREFIX ex: <https://www.linkedin.com/posts/tonyseale_did-you-start-building-your-ontology-as-a-share-7463353332565356545-jY54#>
SELECT ?check ?name ?description
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontology-is-source-code-tony-seale-gpt5-chat-1.ttl> {
    ex:compileGate ex:tests ?check .
    ?check schema:name ?name ;
           schema:description ?description .
  }
}
ORDER BY ?name
Construct Workflow Run live query 
PREFIX ex: <https://www.linkedin.com/posts/tonyseale_did-you-start-building-your-ontology-as-a-share-7463353332565356545-jY54#>
PREFIX schema: <http://schema.org/>
CONSTRUCT {
  ex:ontologyAsCode ?p ?o .
  ?o schema:name ?name .
}
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontology-is-source-code-tony-seale-gpt5-chat-1.ttl> {
    ex:ontologyAsCode ?p ?o .
    OPTIONAL { ?o schema:name ?name }
  }
}
Comment Perspectives Run live query 
PREFIX schema: <http://schema.org/>
PREFIX ex: <https://www.linkedin.com/posts/tonyseale_did-you-start-building-your-ontology-as-a-share-7463353332565356545-jY54#>
SELECT ?comment ?authorName ?topicName ?text
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontology-is-source-code-tony-seale-gpt5-chat-1.ttl> {
    ex:commentThread schema:hasPart ?comment .
    ?comment schema:author ?author ;
             schema:text ?text ;
             schema:about ?topic .
    ?author schema:name ?authorName .
    ?topic schema:name ?topicName .
  }
}
ORDER BY ?authorName ?topicName

FAQ

Questions and answers are named RDF resources.

Why should an ontology be treated as code?

Because it declares organizational meaning using formal logic, relationships, constraints, and axioms; those artifacts need compilation, review, automated checks, and release discipline.

What does it mean for an ontology to compile?

It means syntax is valid, classes are satisfiable, expected entailments are present, contradictions fail the build, and duplicate concepts are detected before release.

Why is plain text important for ontology work?

Plain text gives ontology engineers and LLM agents access to files, diffs, branches, blame, pull requests, and review context.

How should an ontology be governed?

Using the same engineering controls applied to code: branches, pull requests, reviews, automated checks, releases, and builds.

How do LLMs change ontology engineering?

They increase the value of direct source access because an LLM can read, navigate, and edit ontology files like a codebase, instead of only querying an API.

Glossary

Defined terms from the post collection.

Ontology

A formal graph of classes, relationships, constraints, and axioms for a domain.

Ontology as code

Treating ontology source as logic-bearing code that compiles, has tests, and follows engineering governance.

Compile gate

Automated validation that catches invalid syntax, unsatisfiable classes, missing entailments, contradictions, and duplicates.

Plain-text source

Maintaining ontology definitions in files to enable diffs, branches, pull requests, blame, and source review.

HowTo

A compact workflow for putting ontology-as-code discipline into practice.