Raw Statute Schema (kr2akn.raw-law.v1)

This document fixes the intermediate schema produced by kr2akn before AKN emission. It is deliberately independent of any database or server runtime.

1. Input

• Endpoint: Korean Law Open Data API (law.go.kr/DRF/lawService.do)
• Target: eflaw
• Type: XML
• Root: <법령 법령키="...">
• The OC key is read from --oc or LAWGO_OC and is never written to meta.json or the raw snapshot.

A raw snapshot is stored as:

data/raw/eflaw/<law_id>/<YYYYMMDD_HHMMSSZ>/
├── response.xml
└── meta.json      # url (OC removed), fetched_at, bytes, sha256

2. Raw document shape

<법령 법령키="...">
  <기본정보>
    <법령ID>001706</법령ID>
    <법령명_한글>민법</법령명_한글>
    <법종구분 법종구분코드="A0002">법률</법종구분>
    <소관부처 소관부처코드="1270000">법무부</소관부처>
    <공포일자>20260317</공포일자>
    <시행일자>20260317</시행일자>
    <연락부서>
      <부서단위 부서키="...">
        <소관부처명>법무부</소관부처명>
        <부서명>법무심의관실</부서명>
        <부서연락처>02-2110-3164</부서연락처>
      </부서단위>
    </연락부서>
  </기본정보>
  <조문>
    <조문단위>...</조문단위>
  </조문>
  <부칙><부칙단위 부칙키="...">...</부칙단위></부칙>
  <개정문><개정문내용>...</개정문내용></개정문>
  <제개정이유><제개정이유내용>...</제개정이유내용></제개정이유>
  <별표><별표단위 별표키="...">...</별표단위></별표>
</법령>

2.1 조문단위 variants

조문단위 is overloaded. The reader distinguishes by 조문여부 and 조문내용:

The HCONTAINER level is a deliberate escape hatch for raw headings that do not map 1:1 to an AKN hierarchy element. It will map to AKN hcontainer (or level) in the emitter.

3. Intermediate model

RawLaw:

LawMeta:

Unit:

Bylaw:

Annex:

4. Path and label rules

• Path segments use ASCII and are joined with __.
• Headings: part_1, chp_2, sec_9-2, subsec_3, hc_hang_1.
• Articles: art_4, art_4-2 (제4조의2).
• Paragraphs: para_1, para_2 (circled digits normalized to integers).
• Points: point_1, point_5-2 (5의2.).
• Indents: indent_ga, indent_na, indent_ga-2.
• Parent context is always included: part_1__chp_2__sec_1__art_3__para_2__point_1.
• The path is deterministic for a given raw document and is the future eId candidate. AKN wId will use the raw *키 values where available.

5. Corpus check (10 major statutes)

Snapshot date: 2026-10-27. Laws:

Totals:

조문단위 5,037
ARTICLE 4,575
PARAGRAPH 5,799
POINT 1,943
INDENT 18
PART 29 / CHAPTER 181 / SECTION 184 / SUBSECTION 66 / HCONTAINER 2
bylaws 265
annexes 1
warnings 0

Raw field coverage:

조문가지번호 567
조문참고자료 1,332
조문이동이전/이후 35 each
조문변경여부 Y 35 / N 5,002
항번호 circled 5,799
호번호 numeric 1,927 / numeric branch 16
목번호 korean 18

6. Edge cases locked by this schema

1. Heading overload — 조문여부=전문 is used for parts, chapters, sections, subsections, branched sections, and generic 제N항 headings. Body-pattern parsing, not 조문번호, decides the level.
2. Branched headings — 제9절의2 must become sec_9-2, not an article.
3. Empty 항 wrappers — when 항번호/항내용 are empty and the element only contains 호, the points are promoted to the article. This prevents empty AKN paragraphs.
4. Point branches — 5의2. becomes point_5-2; the label is not truncated to 5.
5. Multiple departments — 연락부서/부서단위 is preserved as a list; the top-level 소관부처 is only the representative ministry.
6. Unnumbered annex — 별표번호=0000, 별표가지번호=00 becomes 별표 (not 별표 0), with 별표제목 preserved.
7. Annex tables — 별표내용 may be a box-drawing text table. It is kept verbatim in the intermediate model; the table parser will convert it to AKN <table> in the next phase.

7. AKN emission

The emitter is implemented in src/kr2akn/statute/emit.py; see docs/akn-mapping.md for the full contract.

• PART → part, CHAPTER → chapter, SECTION → section, SUBSECTION → subsection, HCONTAINER → hcontainer.
• ARTICLE → article, PARAGRAPH → paragraph, POINT → point, INDENT → indent.
• Bylaw → hcontainer name="부칙" with article-level children.
• Annex → attachments/attachment/doc/mainBody/table.
• LawMeta → AKN meta/identification FRBR fields.
• Inline citations in main-law text → AKN ref elements.