Sources: corpus forms

Analysis of real fragments: package identifier, act name, the key fragment, form rating (exemplary / debatable / wrong-form choice) and why.

Corpus-wide counts (508 packages; relation-argument shares recounted later):

Corpus-wide counts:

Form	Occurrences	Packages	What it usually is
label labels	121,170	501	declaration names for lawyers
@source anchors	84,702	488	tracing norms to fragments
fragment fragments	44,250	472	addressable text selections
content_hash hashes	45,160	473	pinning texts and bytes
edition editions	1,016	—	expressions of act content
publication publications	919	—	manifestations with bytes
source works	818	—	abstract acts
label official	170,758	476	article and predicate names
label unofficial	29,767	348	author phrasings
label translation	15,726	168	translations
@source(Name) by bare name	≈51,800	488	anchor to an own-package article
@source("urn:…") as a string	≈33,000	—	anchor to a foreign-package fragment
article fragments	13,665 (kz)	472	articles, then paragraphs and points
materialization_status PINNED_UNOFFICIAL_COPY	765	—	copy from the official portal; PINNED_OFFICIAL_BYTES — 196, ABSTRACT_ONLY — 64

OverviewBoundariesCorpus formsPitfalls

Law

1. RK Civil Code — full link (exemplary, inspected)

Package kz.corpus.civilcode — Kazakh Civil Code (Kazakhstan):

Arxo Law

source GK_RK {
    kind code;
    jurisdiction KZ;
    number "268-XIII";
}

Further along the package: a PINNED_UNOFFICIAL_COPY edition, a publication with a uri to the official portal, retrieved_at and a nearby text copy, fragments per article and anchors on every norm. Exemplary: the three identity levels told apart, honest pinning (a portal copy, not PINNED_OFFICIAL_BYTES), bytes nearby.

2. UAE Emiratisation — bare anchor on a fragment (exemplary, inspected)

Package ae.corpus.emiratisation — UAE Emiratisation rules (United Arab Emirates):

Arxo Law

@source(MR455_STRUCTURAL_SCOPE)
rule TwentyToFortyNineSelectedScope(e: Establishment, count: Integer) strict {
    ...
    when employee_count(e, count) and count >= 20 and count <= 49 and selected_economic_activity(e);
    then twenty_to_forty_nine_scope(e);
}

A bare anchor by name — the corpus’s most frequent form (≈51 800 occurrences, 488 packages). Exemplary: the name declared in the package, the anchor traces the norm to the article with no relation attributes.

3. RK Administrative Code — string anchor to another package (exemplary, page data)

Arxo Law

@source("urn:kz:corpus:clir:administrative-code#GK_ART1")

The form for referencing another package’s fragment (≈33 000 occurrences; package kz.corpus.administrative_code, Kazakhstan). Exemplary: an opaque identifier of an external registry written as a quoted string, not a bare name (otherwise LDC-E5103).

4. RK Constitution — morphological forms (exemplary, file verified)

Package kz.constitution.sources — Kazakh Constitution text (Kazakhstan): the only corpus package with forms (label … { genitive "…"; }, 10 occurrences). Exemplary: the nominative is the text field itself, the rest are closed slots of the language family.

Standard and protocol

5. HTTP-framing — one-line pinned text (exemplary, page data)

The one-line text form on a pinned fragment (the same form in three other packages). Exemplary: the grammar allows text_block | string_literal; the choice is readability’s business, the hash counts over the same bytes.

Science and generators

6. The algo.* family — relation attributes by generator (debatable, page data)

relation = formalizes / coverage = partial — 9 351 occurrences each, all in generated algo.* packages, where the generator prints the annotation; formalizes values 9 351, partial 9 305, exact 46. Debatable, not exemplary: attributes exist only where a machine puts them; the corpus’s hand-written packages almost never use them — a practice gap, not exemplary coverage.

7. @source(…) { … } blocks — absent from the corpus (form boundary)

Lexical blocks do not occur in the corpus. Rating: not a choice mistake but an unused form — anchors are put one per norm. The research.sources.lexical_blocks example shows the form works (check OK, law test 2/2).

Teaching packages

8. UZ Civil Code — scenario sets without a publication (exemplary, inspected)

Package uz.civil_code — Uzbek Civil Code (Uzbekistan): 71 sets in [[tests]] (family = "uz.civil_code#authored"), norm texts without pinned bytes. Exemplary for unpinned material: no bytes — no PINNED_* claim, no false LDC-E5206/LDC-E5207.