Early Records Database of the Ainu Language (ERDAL) / アイヌ語古記録データベース / Aynuitak Huskokampi Aomarepu
https://rec.aynu.org

Dataset snapshot of the transcriptions on two platforms:
  みんなで翻刻 (https://app.honkoku.org/projects/ainu): 24 witness volumes, synced 2026-10-17T03:53:32.442Z
  Wikisource (https://ja.wikisource.org/wiki/カテゴリ:アイヌ語): 3 witness volumes, synced 2026-09-13T12:03:48.352Z
Each unit's sources.json record names its platform; pages.jsonl and the TEI headers
link every page to the platform page it was transcribed on.
The Moshiogusa Yukar edition also includes Wikisource revision 224749,
transcribed by mkpoli, with provisional ERDAL phrase readings and translations.

Contents
  source-histories/<source>.json
                            cited histories of composition, witnesses and scholarly editions;
                            bilingual prose, dated events and bibliographic dependencies
  tei/<source>/<unit>.xml   TEI P5 diplomatic transcription, one file per witness volume
  moshiogusa-yukar.json      katakana transcription based on Wikisource, phrase readings, translations and collation
  tei/moshiogusa/yukar-wikisource.xml
                            Wikisource Yukar edition with separate editorial notes;
                            segment numbers follow full stops, not reconstructed verse lines;
                            historical ne/wi allographs are encoded as katakana ネ/ヰ
  pages.jsonl               one record per page; image-based excerpts carry a supplement record; reviewed platform conversions carry a conversion record
                            and everyone who saved it, ranked by characters changed
  entries.jsonl, entries.csv
    gloss and form retain the harvested heading and spelling. JSON glossCorrection
    and correction preserve original, corrected,
    note and id; CSV exposes correctedGloss, glossCorrectionId, glossCorrectionNote,
    correctedForm, correctionId and correctionNote.
    An empty corrected form with a correction ID records an editorial deletion.
    JSON kind="transcription" marks a corrected platform reading, labelled as the
    previous transcription rather than source [sic]. Omitted kind means a source
    error. CSV correctionKind and glossCorrectionKind use transcription/source
    when a correction is present, and are empty otherwise.
                            wordlist items recognised in the transcriptions
                            (gloss, historical form, editorial modern form where present)
  wordlist-coverage.json    unread entry candidates and partially extracted lines,
                            with source, page, line, raw transcription and reason;
                            unit totals exclude headings, notes, wrapped tails,
                            maps, prefaces and running texts. Parallel gloss/form
                            columns count separately as candidate lines.
  places.json               place records, identifications, historical name forms and attestations
  places.geojson            WGS84 locator geometries, with role, precision and coordinate sources
  places-linked.geojson     Linked Places: supported names with locators and dated source citations
  places-linked-coverage.json  included records, omitted attestations and reasons
  place-names.csv           historical spellings with name-origin claims
  place-attestations.csv    witness occurrences linked to names and identification claims
  journeys.json             ordered events, documented stays, evidence, exclusions and legs
  journeys.geojson          events, documented stays and schematic legs; unlocated geometry remains null
  journey-events.csv        event sequence with original dates and manuscript evidence
  research/narita-2010/      visually transcribed research tables, citations, review status
                            and explicit links to manuscript entries; see its README.md
  contributors.csv          transcription credit per contributor: pages, characters changed, page list
  sources.json, sources.yaml
                            source and witness metadata
  hokkai-zuihitsu-entities.json, hokkai-zuihitsu-entities.csv
                            source-specific names and descriptions; JSON includes attestations
                            Editorial descriptions and language/interpretation notes in JSON may
                            contain [[namespace:target|label]] links. CSV uses visible labels.
  hokkai-zuihitsu-mentions.jsonl
                            occurrences with witness, text revision, original form and page/line

Licence
  Transcriptions and their derived data: CC BY-SA 4.0. Attribute to the contributors of the
  みんなで翻刻 project アイヌ関連資料, to the contributors of Japanese Wikisource for the
  units transcribed there, and to ERDAL (https://rec.aynu.org).
  Research exports carry their own source attribution and rights metadata.
  The Yukar edition credits mkpoli and the Wikisource contributors under CC BY-SA 4.0;
  its JSON and TEI include the permanent revision URL and contributor history.
  Facsimile images are not included; each source's witness record names the holding
  institution and the terms under which it publishes its images.

Editorial status
  Modern-form interpretations carry a certainty (high / medium / low) and a citation.
  basis is set when a program proposed the reading: machine (not matched to a dictionary)
  or machine-checked (every word found in a dictionary). It is empty otherwise.
  morphemes is a reading's morpheme analysis after the Leipzig Glossing Rules: "-" between
  morphemes, "=" at a clitic, "~" in reduplication (ko-perke-p). It may give the analysed
  shape of a word rather than its spelling (sir-ratci for sinratci). morphemeGloss, when
  set, glosses each morpheme with the same words and boundaries (APPL-break-NMLZ).
  In JSON a citation links each work it names as [[db:<slug>|label]], the slug being the
  work's record at https://db.aynu.org/sources/<slug>; entries.csv uses the visible labels.
  A form without a modern field has not been interpreted.
