# Open-source dictionaries bundled with this app

This app is a **reader**: it ships with open-licensed dictionary data only, and
loads further dictionaries only from files the user provides. Attribution below
covers everything bundled in the `samples/` folder of a public deployment.

| File | Source | License | Attribution |
|---|---|---|---|
| `samples/wordnet31.mdx` | WordNet 3.1, Princeton University; **plus** IPA from CMUdict (public domain / BSD) and from Simple English Wiktionary (CC BY-SA) | WordNet 3.1 license for definitions; the Wiktionary-derived IPA portions are shared under CC BY-SA | "This product uses WordNet 3.1 (Princeton University, https://wordnet.princeton.edu)" — citing G. A. Miller (1995), *WordNet: A Lexical Database for English*, Communications of the ACM 38(11), and/or C. Fellbaum (1998), *WordNet: An Electronic Lexical Database*, MIT Press |
| `samples/simple-en.mdx` | Simple English Wiktionary (derived content) | CC BY-SA 3.0+/4.0 (dual-licensed) & GFDL | "Dictionary content from Simple English Wiktionary, CC BY-SA (https://simple.wiktionary.org)" |
| `samples/sample.mdx` / `.mdd` | Generated by this app for testing | Same license as the app | — |

If you add another open dictionary (e.g. GCIDE, a full Wiktionary build), add a
row here **and** to the "Dictionaries & licenses" section in the About panel
(`index.html`) before deploying. CC BY-SA content requires visible attribution.

Commercial dictionaries (OALD, LDOCE, …) are **never** bundled or distributed by
the public deployment — users load their own files locally via ⚙ Advanced.

## EN–VI dictionary

The optional English–Vietnamese dictionary ("Từ điển Anh–Việt v1.1", 386,600
entries incl. inflected forms) is built and maintained by the **Open Vietnamese
Dictionary Project** (OVDP) and distributed via catusf/tudien — MIT license.
Source: https://github.com/catusf/tudien

## Word frequency data (English)

- `data/freq-en.json` — top 10,000 English word frequency bands derived from
  [hermitdave/FrequencyWords](https://github.com/hermitdave/FrequencyWords)
  (OpenSubtitles 2018 en_50k), MIT License. Bands: top 1k / 2k / 5k / 10k.

## Pronunciation (IPA) data — `data/ipa-en.json`

Headword → IPA lookup table used by the pronunciation strip (English
pronunciations only — no definitions are shipped from these sources):

- Primary: [thichhoc-org/thichhoc-dict](https://github.com/thichhoc-org/thichhoc-dict)
  release v0.9.0 (2026-08-05) — dictionary data **CC BY-SA 4.0** (required by the
  share-alike Wiktionary upstream; full attribution chain in that repo's
  `dict-en-vi/LICENSE-DATA` and `ATTRIBUTION.txt`). Only the `<span class="ipa">`
  pronunciation field is extracted; upstream sources are WordNet 3.1, CMUdict and
  Wiktionary.
- Fill for function words missing there: [cmusphinx/cmudict](https://github.com/cmusphinx/cmudict)
  (CMUdict, public domain / BSD), converted with the standard ARPAbet→IPA mapping.
- Built by `tools/build-ipa-json.py` (in the workspace, not deployed); the JSON is
  lazy-loaded on first use and served as-is.

## Learning-set frame data — `data/ngsl.json`

The "learning set" (nhóm "Học chính" on the 2,000-word pages) is framed by the
**New General Service List (NGSL) 1.2** by Charles Browne, Brent Culligan and
Joseph Phillips — **CC BY-SA 4.0** (this redistribution therefore carries the
same license; see https://www.newgeneralservicelist.com/new-general-service-list):

- `core`: 2,809 lemmas with official SFI ranks, from the official
  `NGSL_12_stats.csv` (downloaded 2026-10-01 from newgeneralservicelist.com).
- `supplementary`: 52 weekday/month/number lemmas from the official
  `SUP_lemmatized.csv`.
- Used by `tools/build-word-pages.cjs` and `tools/classify-wordlist.cjs` to
  label each word page: core (lemma in NGSL) / inflection (dạng chia, lemma
  shown) / off-list (fragment, abbreviation, proper noun, or lemma outside
  NGSL) / vulgar (separate warning group). No claim is made that a word is
  "top 2,000 by some list" — frequency bands come from OpenSubtitles, learning
  membership from NGSL, and both are labeled distinctly.

## Academic Word List — `data/awl.json`

The "Học thuật (AWL)" label and the supplement pages of the word-pages set
(v95) come from the **Academic Word List (AWL)** — 570 word families by
**Averil Coxhead**, Victoria University of Wellington (2000), distributed free
for educational/research use with this attribution (see
https://www.wgtn.ac.nz/lals/resources/academicwordlist/). Word + sublist
assignment imported from the Simple English Wiktionary rendering of the list;
3 official Sublist-9 members (protocol, founded, duration) restored from the
published list. Used by `tools/build-word-pages.cjs` (batch 2) and the
"Thuộc AWL" hub filter. AWL membership is labeled distinctly and never
conflated with OpenSubtitles frequency bands or NGSL learning membership.

## Spoken audio — `audio/us/*.m4a`, `audio/gb/*.m4a` (P1, 01/11/2026)

Per-word pronunciation audio is pre-rendered with **Piper TTS** (open-source
neural voices, https://github.com/rhasspy/piper) and hosted on-site:

- US voice: `en_US-amy-medium` (Amy, American English) — `audio/us/`
- GB voice: `en_GB-alba-medium` (Alba, British English) — `audio/gb/`
- Piper voices are free for commercial and non-commercial use (see
  https://github.com/rhasspy/piper-voices for per-voice training-data licenses).
- Script: `tools/render-audio.py` (voices live OUTSIDE the repo at
  `~/piper-voices/`, never committed). Files served with `preload="none"` —
  no bandwidth until the user taps play.

## Anki .apkg export (v119 A3)

- `js/vendor/sql-wasm.js` + `js/vendor/sql-wasm.wasm` — [sql.js](https://sql.js.org) 1.13.0 (SQLite compiled to WebAssembly), **MIT License** (full text: `js/vendor/sql-js-LICENSE.txt`). Loaded lazily on first .apkg export, cached by the service worker after that.
- `js/vendor/apkg-writer.js` — original code for this project (zip STORE writer + Anki 2 schema v11 builder), no third-party zip/deflate code.
