# Atlas of Australian Languages — open data downloads

**Data release 1.1.0** · build 2026-07-12 · generated by `apps/web/scripts/atlas/build-exports.ts`

These files are a deterministic, reproducible snapshot of the atlas, built from
the committed static artifacts under `apps/web/data/atlas/` (which are in turn
joined from the upstream datasets listed below). No blank ever passes as a
value: every missing datum is an explicit honest state (`none`, `unknown`,
`not_profiled`, `undated`), and coordinates carry a provenance flag.

## Files

| File | Rows | What it is |
|------|------|------------|
| `languages.csv` | 980 | One row per genuine languoid: identity, family, coordinates + provenance, tier, endangerment, grammar/lexicon coverage, deep-time divergence and full classification chain. |
| `languages.geojson` | 809 | GeoJSON FeatureCollection of the 809 located languages (Point geometry); the 171 unlocated languoids are honestly omitted, never given a fake point. |
| `grammar-values.csv` | 30351 | Long-format grammatical feature codings (one row per language × feature). value/present/absent/unknown(?) states kept distinct — read the state column; unknown = not recorded. |
| `theses.json` | 8 | The 8 "why did the languages move?" scholarly theses, contradiction-preserving — each with mechanism, proponents, evidence for AND against, hard facts, contestation level and full citations (with DOIs). |
| `sources.bib` | 56 | BibTeX for the upstream datasets and the 48 scholarly references behind the movement theses (DOIs as published). Includes an entry for citing the atlas itself. |
| `CITATION.cff` | — | Citation File Format metadata for the atlas itself (title, authors, version, licence, release date, upstream references) — machine-readable "cite this". |
| `README.md` | — | This data dictionary. |

## Data dictionary — languages.csv

One row per **genuine languoid** (980 of 1029 registry nodes; the 49 Sign/Pidgin/Bookkeeping/mixed nodes are appendixed, not counted). Mi'gmaq is excluded as non-Australian.

- **slug** — stable atlas id (prefers glottocode; falls back to AUSTLANG code / DB slug). The canonical URL is `/atlas/<slug>`.
- **name** — canonical (English-catalogue) name. Autonyms are shown on the profile as *unverified candidates* only, so are not a CSV column.
- **family** — top-level family; `unclassified` where the open exports give no classification.
- **macroarea** — Glottolog macroarea.
- **glottocode**, **iso639_3** — codes where present (blank = none in the open export).
- **austlang** — AIATSIS AUSTLANG code(s), `;`-joined (a languoid may map to several).
- **lat**, **lon** — coordinate if located; blank if location unknown.
- **coord_provenance** — `glottolog` (real point) · `austlang` (approximate) · `derived_centroid` (mean of sibling leaves, a drawing convenience) · `none` (unlocated — never given a fake point).
- **coord_approximate** — `true` for austlang/derived points; blank when unlocated.
- **tier** — coverage tier: comprehensive / documented / classified / listed.
- **endangerment** — Glottolog AES label (`unknown` where not coded).
- **grammar_profiled** — whether a Grambank/WALS/AUS feature profile exists.
- **grammar_feature_count** — number of *recorded* (non-`?`) feature codings for the language.
- **lexicon_state** — `live` (browsable words in DB) · `open_resource` · `pointer_only` (rights-managed catalogue link) · `none`.
- **word_count** — live DB word count (blank unless `lexicon_state=live`).
- **deep_time_divergence_bp** — divergence age (years BP) of the language's lineage from the Bouckaert 2018 tree, where dated. **This dates a LANGUAGE lineage, not a population arrival.**
- **classification_chain** — full Glottolog path, ` > `-joined.

## Data dictionary — languages.geojson

FeatureCollection of the **809 located** languages (Point geometry, `[lon, lat]`). Properties mirror the key CSV fields plus `coord_approximate`. Unlocated languoids are **omitted** (see `metadata.unlocated_omitted`) — they remain in `languages.csv` with `coord_provenance=none`. No coordinate is ever fabricated.

## Data dictionary — grammar-values.csv

Long format: one row per language × feature coding.
- **language_slug**, **language_name** — the language.
- **feature_id** — Grambank (`GBxxx`) / WALS / AUS-extension feature id.
- **source** — `Grambank` / `WALS` / `AUS extension`.
- **layer**, **domain** — grouping metadata.
- **gloss** — the feature's plain-English description.
- **value** — the coded value.
- **state** — `value`/`present`/`absent` (real codings) vs `unknown` (`?`, not recorded). These are kept **distinct** — an absent (0) is a real datum; an unknown is a gap.
- **raw** — the original coded token.

> `grammar_recorded_agreement` (used in the Grammar lens) is agreement over *jointly-recorded* Grambank features only — **not** overall grammatical similarity and **not** genetic relatedness. Always read `n_joint`.

## Licences — attribute accordingly

These are **open data under the upstream licences**. The atlas's own *derived* layer (the joins, tiers, provenance flags, coverage tables) is released **CC-BY-4.0**. Individual fields inherit their source licence:

- **Glottolog** (5.3 CLDF) — **CC-BY-4.0** — classification, coordinates, ISO/glottocode, AES endangerment
- **Grambank** (v1.0.3 CLDF) — **CC-BY-4.0** — grammatical feature profiles (195 features)
- **WALS** (2020.4 CLDF) — **CC-BY-4.0** — supplementary grammatical features (coded values baked into grammar.features)
- **Australianist typology extension (AUS extension)** (mobtranslate-pg typology plane) — **CC-BY-4.0 (derived)** — Australia-specific grammatical features + recorded-agreement similarity (grammar.features, related[])
- **AIATSIS AUSTLANG** (data.gov.au export) — **CC-BY-4.0** — canonical AU language codes, approx. coordinates, alt-names
- **Bouckaert, Bowern & Atkinson 2018 (Phlorest)** (phlorest CLDF) — **CC-BY-4.0** — dated Pama-Nyungan phylogeographic tree (deep_time)
- **PHOIBLE** (referenced) — **CC-BY-SA-3.0** — phonological inventories (pointer; not ingested in P0)
- **E. M. Curr, The Australian Race (1886-87)** (archive.org OCR) — **Public Domain** — historical locality wordlists (live DB curr_* + appendix)
- **English Wiktionary** (kaikki.org export) — **CC-BY-SA-4.0** — open lexical resources for some languoids
- **mobtranslate-pg (live DB)** (read-only snapshot @ 2026-07-12T00:00:00Z) — **mixed (see per-source)** — live lexical word-counts

**Share-alike note:** Wiktionary-derived lexical data is **CC-BY-SA-4.0** and PHOIBLE is **CC-BY-SA-3.0**; if you redistribute those fields, keep the share-alike terms. Everything else is CC-BY-family (attribution only).

**CC-BY vs CC-BY-NC note:** the Bouckaert et al. 2018 phylogeny is used here via the **Phlorest CLDF (CC-BY-4.0)**. The same tree is also distributed through **D-PLACE**, parts of which carry **CC-BY-NC** — we deliberately source the CC-BY-4.0 Phlorest release so the deep-time layer is not encumbered by a non-commercial clause.

## How to cite

**Suggested citation**

> Mob Translate project & Claude (2026). The Atlas of Australian Languages, data release 1.1.0. https://mobtranslate.com/atlas (accessed <date>).

Plus cite the upstream datasets you rely on (see `sources.bib` / `CITATION.cff`). A copy-pasteable BibTeX block and the machine-readable `CITATION.cff` are in this folder.

## Honesty & uncertainty

- Coverage is reported as fractions, never rounded to imply completeness: 203/980 grammatically profiled; 224/980 with a dated tree position; 809/980 located.
- Deep-time ages are the **language lineage**, with weak posterior support at the deepest Pama-Nyungan nodes — not a settled fact and not a population-arrival date.
- Autonyms are unverified candidates from alt-name lists; they are not asserted here.
- The movement theses (`theses.json`) are held contradiction-preservingly: every thesis carries mandatory evidence *against* and a contestation level; no single "winner" is declared.

_Communities are the final authority on their own languages. Rights-managed materials (AIATSIS / language-centre dictionaries) are catalogue pointers only — never scraped. To correct or request removal of any record, contact ajax@mobtranslate.com._