Source readiness
Start with the legal sources you have.
LawVM defines an evidence-preserving path from source inventory to structured legal units, amendment relationships, bounded transition checks, and point-in-time state. For scans, document reconstruction preserves uncertainty alongside the proposed text.
Beyond transcription
A scan can become a reviewable evidence package.
A high-assurance recovery workflow can preserve the source, account for every page and region, compare independent observations, retain alternatives and unread material, and emit a deterministic candidate derivative for review. Unsupported units remain unresolved instead of being silently guessed.
Source and page account
Freeze the supplied bytes and package positions; inventory page images, embedded objects, locators, and missing or unreadable material without inventing page artifacts.
Separate evidence lanes
Keep embedded text, OCR, pixels, geometry, layout, source-native structure, and external transcriptions distinct. Agreement is evidence; disagreement is a review route, not an automatic winner.
Claim-specific reconstruction
Characters, reading order, grouping, headings, tables, footnotes, and coverage are different propositions. A strong text result cannot hide unresolved order or structure.
Derivative and review boundary
Bind every candidate unit back to source loci, independently reread the output, preserve alternatives and residuals, and promote only the units allowed by the declared review profile.
Maturity ladder
What can be produced now?
| Starting point | First supportable output | Additional evidence required |
|---|---|---|
| Scanned gazettes | Exact source/package and page-image account; coordinate-bound visual and OCR observations; explicit unread, ambiguous, and unclaimed regions | Human validation of characters, order, grouping, tables, and legal structure before historical consolidation |
| Searchable PDFs | Preserved manifestations; raw text and geometry observations; page/region accounting; candidate lines, structure, citations, alternatives, and residuals | Review of logical order, documentary structure, and amendment semantics |
| HTML or partially structured XML | Source trees, identity inventory, structure diagnostics, unknown-element reports | Amendment sources for historical replay |
| Original and amending acts | Dependency inventory, clause/payload surfaces, typed operation candidates | Independent dry-run evidence before replay promotion |
| Before/after versions | Transition bundles, mutation-boundary dry-runs, typed agreement/residual rows | Source review and corpus accounting for correctness and coverage |
| Stable API, XML, AKN, or ELI | Reproducible acquisition, version/effect graphs, bounded replay and conformance work | Legal review for interpretation above the text-state layer |
For legal-information institutes, archives, and publishers
Source evidence can be produced before replay exists.
Japan
e-Gov history, revision XML, bulk XML, attachments, enacted texts, and a loss-preserving source AST create a source/PIT census without claiming executable amendments.
Poland
ELI/Sejm discovery, official PDFs, HTML, structure, references, and source accounts establish what material exists before Polish clause semantics are implemented.
South Korea
Typed source-lane and acquisition accounting records missing configuration, endpoint gaps, skips, and transport failures without treating them as an empty corpus.
Switzerland
Fedlex impacts and multilingual manifestations can be inventoried precisely while operation lowering and replay remain explicitly unauthorized.
First bounded assessment
Return a readiness map with next requirements.
Inputs
A representative source sample, publication registers, identifier and citation conventions, known authority rules, and local source expertise.
Measurements
Package/page/image coverage; text, geometry, layout, and structure observations; alternatives and residual regions; independent readback; identity, version linkage, amendment closure, dates, and expressions.
Outputs
Source account, claim-by-claim evidence map, achievable citation granularity, reviewed candidate derivative where supported, blocked-source map, transition feasibility, and next integration boundary.
Decision
What can run immediately, what becomes possible with named additional sources, and what remains unavailable.