Metadoc — preface-front-matter-enrichment-use-cases.md
Created: 24-07-2026 · Last updated: 05-09-2026
| Field | Value |
|---|
| Subject | preface-front-matter-enrichment-use-cases.md |
| Purpose | Proposal: what OCR front matter teaches and which use cases it enriches |
| Audience | Maintainers, DH pipeline owners, paper authors |
| Provenance | Session proposal after H1560–H1580 · Grok 4.5 (grok-4.5) |
| Status | Living proposal (not a locked roadmap) |
Ranked improvement backlog
Promote UC-3 structured legend store to a committed generator with schema. Done (H1591).Wire UC-1 tooltips from legend store into a site component. Done (H1593) for PWG/PW.UC-11 FAIR path map on ocr-prefaces + reciprocal links. Done (H1595).- UC-2 pilot: PWG↔PW(↔AP90) work-identity table from pref expansions + body counts — MW
struck by H1854 M1, root cause investigated and confirmed by H2279 (06-08-2026):
its p. xxxiii+ "List of Works and Authors" legend was never OCR'd upstream —
pref_abbr_crosscheck.py --dict MW parses 0 keys by design (documented in the script
header), not a parser defect. Re-add after that page is transcribed into MWS/prefaces/.
- Resolve open questions (rare/MS UI; Zenodo DOI at next dict release).
H1854 hostile-read verdict (04-08-2026)
Adversarial pass over every claim and P1 row of the subject
(H1854,
Fable 5 claude-fable-5). Default posture REFUTED: a row survives only if a committed
artifact it would consume exists on disk. Verdict: the proposal is sound; 3 Major and
4 Minor findings; both remaining P1 rows survive narrowed, one P1 row was already shipped.
Major findings
| # | Lands on | Finding |
|---|
| M1 | use-cases §UC-2, line 92 + backlog item 4 below | UC-2's named input "per-dict legend tables (PWG, PW, MW, AP90)" has no committed MW artifact: scripts/out/ contains no mw_* file at all (no legend, no crosscheck, no decompose). MW prefaces are OCR'd (MWS/prefaces/) but the legend was never parsed. MW struck from the UC-2 pilot until an MW legend emit exists; the pilot that can run today is PWG↔PW (from scripts/out/pwg_legend.json + pw_legend.json) extended by AP90 via ap90_pref_abbr_crosscheck.tsv. |
| M2 | use-cases § Priority map, line 208 | The P1 row still lists UC-4, which shipped 28-07-2026 (H1760 status line at UC-4 itself, line 112). Stale: the real P1 build decision is UC-2 + UC-6 only. Fixed in the subject in this same pass. |
| M3 | use-cases §UC-6, lines 122–128 | UC-6's headline output "legend coverage of corpus citations" already exists: GH-4 is Tested — 95.3 % of corpus <ls> citations fall in legend-documented dictionaries (1,187,169 / 1,245,644; guides-hypotheses.md line 40). The row survives only for what is genuinely missing: the per-dictionary missing-legend vs unused-legend split, which no committed artifact yet computes. |
Minor findings
| # | Lands on | Finding |
|---|
| m1 | use-cases §UC-3, line 102 | "JSON/TSV per dict" overstates the legend store: *_legend.json exists for 2 dicts (PWG, PW — src/data/pref-legends.json self-declares pilot), while crosscheck/decompose TSVs cover ~20. Accurate as "crosscheck per dict, legend store pilot-only". |
| m2 | this metadoc, header + revision history | Metadoc had drifted: Last updated 24-07 against a subject updated 28-07; revision history was missing the H1592 (UC-8 fold table) and H1760 (UC-4 reading packs) ship rows. Header bumped this pass. |
| m3 | use-cases §UC-9, line 150 | "printed addenda/corrigenda … in front matter" — no committed addenda-page artifact was located in any prefaces/ tree this pass. P2 row keeps its hedge, but its input is unverified, not confirmed. |
| m4 | use-cases § Related, line 6 | Issue csl-guides#123 could not be verified this session (gh unavailable offline). Noted, not adjudicated. |
P1 adjudication (acceptance: named committed artifact or struck)
Claims confirmed against disk (spot record)
Six residual classes exactly as named (ortho, rare, ocr_key, spacing, true_unused,
ambiguous — pref_only_decompose.py);
UC-8 fold table 26 examples + --self-check real
(pref_fold_table.json,
pref_key_body_align.py lines 20/331);
UC-11 chain complete (PWG METHODS ·
PW METHODS ·
pref_legend_emit.py ·
ocr-prefaces ·
naming authority);
UC-13 homes exist (residual methods appendix ·
ALL_pref_only_decompose.md);
UC-14 tri-lingual inputs exist for PWG (27 pp DE/EN/RU), PW, MW (*pref_all.{en,ru}.md);
all named site consumers exist (abbreviations-comparison, citation-graph, corpus-attestation,
entry-anatomy). Grammar-metalanguage claim (§5) backed by committed mw72/md/shs/cae
crosscheck outputs.
Limitations
- Use cases are framed, not implemented (except P0 partials already in tools/site).
- Residual counts age; re-run census tools for paper-grade numbers.
- Rights for some front-matter repos may differ; UC-12 still needs publish-safety.
Revision history
| Date | Change |
|---|
| 24-07-2026 | Initial proposal + 14 use cases + priority map |
| 24-07-2026 | UC-11 concrete path map (H1595 FAIR cross-links) |
| 04-08-2026 | H1854 hostile-read verdict: 3 Major / 4 Minor; UC-2 + UC-6 survive narrowed, UC-4 confirmed shipped; priority map fixed in subject. Fable 5 (claude-fable-5) |
| 06-08-2026 | UC-2 MW investigated (H2279): blocked on missing p. xxxiii+ legend OCR, documented delta, not a parser fix |
Dr. Mārcis Gasūns