Markdown for LLMs

Definitions: corpus forms

The source Markdown for this article. Copy it into your assistant or download it as a text file.

# Definitions: corpus forms

Breakdown of real fragments: package identifier, human act name, form rating
(exemplary / debatable / wrong form choice) and why.

## Science

### 1. NASEM reproducibility — public exact with three conjuncts (exemplary)

Package `nasem.reproducibility` (NASEM reproducibility definitions, US science):

```law
pub definition ReproductionSetupMatchesDefinition(att: ReproductionAttempt) exact {
    label ru unofficial "Постановка попытки отвечает определению воспроизводимости: те же данные, тот же код, те же условия анализа";
    when used_same_input_data(att) and used_same_code_and_methods(att) and used_same_conditions_of_analysis(att);
}
```

Exemplary: a named concept referenced by other norms of the package (the
file’s following definitions answer the same question for neighbouring
concepts); `pub` exports the concept; the source anchor
traces to the fragment. The body is a conjunction of three literals: the
sufficient half infers the concept only when all three hold at once.

### 2. Same file — family of definitions on one question (exemplary)

The same package holds `ReplicationSetupMatchesDefinition` (own data, same
question) and `GeneralizabilityClaim` (two arguments, one
condition). Exemplary as a method: one
domain question — several named concepts side by side, each with its own
condition; the consumer picks the name instead of rewriting conditions. The
reverse would be the mistake: one concept with a disjunction of unrelated
conditions inside.

## Standards

### 3. W3C PROV — definition over a same-named relation (exemplary)

Package `w3c.prov` (W3C PROV, provenance standard), entities and activities:

```law
pub definition Communication(informed: ProvObject, informant: ProvObject) exact {
    label ru unofficial "Сообщение: одна деятельность использовала сущность, порождённую другой, и сама сущность не названа";
    when communication(informed, informant);
}
```

Exemplary and instructive: the definition name (`Communication`) differs in
case from the relation name (`communication`) — the namespaces do not
collide, and the concept names a pattern over the fact. Lowercase is the
fact, uppercase the concept: a naming discipline that removes the “why no
`LDC-E1201`” question.

## Law

### 4. Civil Code of Kazakhstan — qualification as a strict rule instead of a definition (debatable)

Package `kz.corpus.civilcode` (Civil Code of Kazakhstan), natural persons — qualification
of a person as a natural person is written as `rule … strict`, not as
`definition`. Debatable, not a mistake: observed behaviour is the same, but
the proof-graph explanation says “by rule”, not “by definition”, and the
`was_derived_from` edge is absent. A possible reason is the qualification
being woven into the article: when the condition and the concept live in one
sentence of the act, the author picks a rule; a definition pays off where
the name is reused. Corpus-wide counts record the same practice
quantitatively: `rule … strict` in hundreds of packages, typical use —
definitions, qualifications, structural consequences (in detail — the
[strict-rules corpus forms](/constructs/rule-strict/corpus-forms/)).

## Teaching packages

### 5. This section’s examples (runs green)

Package `research.definition.exact_minor` — `definition minor … exact` without a
separate `relation`: `check OK`, `law test` 2/2. Fixes the main formality:
the relation is synthesized, redeclaration is `LDC-E1201`.
Package `research.definition.strict_classification` — `classification driver …
strict`: `check OK`, `law test` 2/2, observed behaviour matches the
definition.