Course outline
AcademyModule 4 · Domain vocabulary and decision records
Ontology: manage vocabulary, types, and properties
What you'll learn
- Distinguish derived vocabulary from your declared schema
- Understand the difference between Auto-Tune saving and accepting schema proposals
- Apply vocabulary changes to older notes and preserve source material with decisions
You can start constructing an ontology without hand-authoring a schema from scratch. Consilience derives vocabulary from documents and proposes custom kinds and properties. A domain owner reviews meaning, edits it, and approves changes.
Normalize expressions with standard vocabulary
“Built”, “developed”, and “created” may describe the same relationship. Canonical predicates and their variants bring later extractions into a consistent language. Custom entity kinds can refine the built-in kinds for your work.
Define in Ontology; inspect in Entity types
The graph sidebar’s Ontology section manages the domain, reading persona, kinds, relations, and properties. Entity types shows the graph’s actual kind distribution and untyped entities, with controls for correcting assignments.
| Setting | What it defines |
|---|---|
| Domain and persona | Your field and the perspective for reading sources |
| Custom kinds | A name, description, and core parent kind |
| Relations | Canonical predicates, descriptions, and variants |
| Properties | Names, data types, and applicable kinds for literal values |
Auto-Tune and schema proposals
Auto-Tune samples documents, derives the domain, reading persona, and predicates, and saves the result. It preserves existing custom kinds and properties. It is a write operation, so read the execution notice before starting.
Review kinds and properties before saving
The kind and property proposal buttons return candidates without saving the schema. Accept or edit useful candidates, then save in Ontology. The domain owner decides names, meanings, and where a property applies.
AI judgment and human correction
AI judges candidates as same, broader, narrower, related, or distinct. Identity merges proposed only from graph structure wait for human confirmation. Other structural relationships follow the automatic-add setting. Correct mistakes in the Synonyms and Relationships sections.
Apply new vocabulary to older documents
Saved vocabulary is used on the next extraction. Re-extract previously processed notes to apply a new policy. When the document itself changes, enabling automatic extraction on save processes the changed note again. That desktop setting is off by default.
New note: “Acorn built Orbit.”
Canonical predicate: develops
Variants: built, developed
Next extraction: Acorn → develops → OrbitTune vocabulary with your own material
- Import a few documents and open Ontology.
- Read the Auto-Tune execution and save notice, then inspect its result.
- Request custom kinds or properties, choose useful candidates, and save.
- Re-extract one existing document and check its types, relations, and sources.
You can compare the intended meaning with the graph’s actual output and original evidence.
Preserve the source and the decisions
Source notes remain editable Markdown. Search indexes can be rebuilt, but vocabulary and human merge or split decisions cannot all be recovered from the source text alone. Export the graph, vocabulary, and decisions when moving the workspace.