How FCOIM relates to conceptual, logical & physical data modeling (and more)
“Conceptual / logical / physical” usually names the three levels of data modeling. A fact-based conceptual model is a different thing — the single, natural-language source those levels are generated from.
Distinction → vertical architecture (legal source → concepts) → horizontal method → artifacts → governance & lineage → the semantic layer & the LLM.
First, the distinction
Two meanings hide behind the same three words. The three levels describe how abstract a schema is. The fact-based conceptual model is not one of those boxes — it sits before them and produces them.
Not boxes drawn first — concrete examples verbalized as sentences, validated with domain experts. Communication-oriented, attribute-free, and more precise than a traditional conceptual data model.
one elementary fact, with a concrete population behind it
The vertical architecture — where the facts & definitions come from
Before the horizontal method runs, a vertical architecture grounds it: it descends from the authoritative legal source down to structured concepts.
The vertical architecture is provenance; the FCOIM method is process. They meet where the BOM’s concepts & containers become the object and fact types of the Information Grammar.
FCOIM as the generative source — the horizontal method
The stepwise method turns those grounded facts into a validated, machine-processable model — the Information Grammar — from which every downstream artifact is generated.
Capture facts
Collect concrete examples; verbalize them as natural-language sentences.
Verbalization & concrete examplesClassify
Group sentences into fact types; identify object types & label types.
ClassificationQualify
Name the roles each object plays; fix populations & readings.
QualificationConstrain
Add uniqueness, totality, set, value & subtype constraints.
Constraint captureValidate & verify
Re-verbalize to experts; test constraints against real populations.
Validation cycleGenerate
Transform the grammar (grouping) into models & artifacts — deterministically.
GenerationOne model, many artifacts
Because the Information Grammar is complete and formal, the same validated model generates the three data-modeling levels and semantic / interchange outputs — all from one source, so they never drift apart.
FCOIM Information Grammar
Validated elementary facts + constraints — the single source of definition
Governance from the source catalog & end-to-end lineage
Map your existing source data catalog onto the FCOIM model, and governance stops being a separate spreadsheet: definitions, ownership and quality rules attach to fact types and flow all the way down to physical columns.
Source data catalog
- Source systems, tables & columns
- Existing data dictionary terms
- Owners, sensitivity, quality rules
FCOIM model + governance
- Catalog columns mapped to fact types
- Definitions, ownership & rules bound to concepts
- Glossary tied to real, verbalized facts
Governed models
- Conceptual → Logical → Physical
- Glossary, JSON, OWL, XSD…
- Every element traceable to a fact
Any physical column all the way back to the legal article that requires it — and forward again.
A term means the same in glossary, ERD & database. No drift.
Change a fact once; see every downstream artifact before regenerating.
In today’s terms — the semantic layer & the LLM
Two things everyone is asking about map cleanly onto this architecture — one you already have, and one to keep on a leash.
Reads legal articles; drafts annotations, candidate terms & facts. It accelerates the vertical architecture — but every suggestion runs through validate & verify. The LLM proposes; it never defines.
The semantic layer grounds the LLM: the glossary, ontology & constraints become its retrieval context, so answers stay consistent with the agreed definitions instead of hallucinating them.
Download 
Use Online