Elucidário Madeirense: technical documents

These documents describe how the modern multilingual edition was built. They are kept for publication on the website as technical notes. Machine-readable reference tables live in kb/.

Start here

Document	Contents

Translation standards

Document	Contents
style/README.md	How the guides combine into translation prompts; the QA checklist
style/core.md	Shared rules for all languages: fidelity, register, block types, numbers and old currency, tables, names, headwords, termbase
style/en-GB.md, de, fr, it, hu, nl, uk, ru	Per-language style guides
style/DECISIONS.md	Decisions confirmed by the editor
transcription_uk.md, transcription_ru.md	Portuguese → Ukrainian/Russian name transcription: religious titles, historical figures, and historical/legal terms
translation_strategy.md	Model pilot: quality scores and the chosen strategy

Reference tables (kb/)

File	Contents
../kb/taxonomy.yaml	Article-type taxonomy v1 (12 classes, 90 subtypes, person-role facet)
../kb/termbase.yaml	1,195 Madeira/Portugal-specific terms with policy and renderings in 8 languages
../kb/religious_titles.yaml	267 Marian, Christological and saint titles with established equivalents and sources
../kb/historical_figures.yaml	374 historical figures with established names per language (Wikidata-verified)
../kb/names/	Per-language name tables (persons, places, institutions)
../kb/names_seed_ru_uk.jsonl	Authoritative worked examples for the Cyrillic standards

Pipeline reports

Report	Contents
reports/phase2.md	Layout extraction and article segmentation (hidden headwords, sub-articles)
reports/phase3.md	OCR correction: rules, LLM passes, review
reports/taxonomy.md	Taxonomy design
reports/jev_benchmark.md	Jev vs Claude on classification tasks