open-standards/aac/aac-rs/tools/pagesets/phrases/reports/translation-status.md
Transaction Science 7c29c52054 feat(aac pagesets): four phrase page sets, built from readable sources, in English and ten languages
Our own page sets for the reference app and any Open Board Format app:

- Adult conversation (41 boards): an always-row (Back, Yes, No, I don't
  know, Wait), a 0-10 scale, pain, comfort and feelings scales, repairs,
  word finding, six scripts as one-press-per-line sequences, cards that
  explain aphasia, ALS/MND, PPA, brain injury and Parkinson's in the
  person's own voice, care, positioning and breathing requests that
  claim nothing is sent, life-story and people templates for the
  person's photos, a page for banked messages, and a staged plan for PPA
  written for partners.
- Intent first in 12, 20 and 40 cells: start pages by purpose, predictable
  branches, operational commands in one place on every page, popups that
  close after a choice, and page numbers for a printed partner book.
- Themed phrases: nine 6x3 pages of starters, words and messages.
- First switches and messages: one-message pages, sequences, press for
  fun, two and four choices, and learn-to-scan pages.

tools/pagesets/phrases/build.py validates the YAML sources (layout,
links, symbol concepts, proprietary names, medical-claim wording) and
writes the packages, a catalogue and the licence into --out DIR,
reproducibly; --check fails on a non-reproducible build or stale reports.
Packages are generated by the site build, not committed.

Every text was drafted, back-translated by a reader who never saw the
English, and judged, in up to three rounds: 10,748 of 10,750 agreed
(es fr de pt it zh ja ar hi ru), recorded per text as "agreed by two model
readers; native review pending"; the two others stay in English, marked.

A new aac-core test runs the generator, then imports every package with
the app's importer and an independent OBF validator, and checks structure,
links, symbol references, that no language moves a cell, that each cell is
in one language and says so, and that translated cells import as the
person's own words with their chosen symbols.

The design notes propose (not implemented) a sequence extension, popups,
speak-only pages, links that say words and one-target scanning, with SPEC
text, and how the chooser should offer each set.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-10-05 12:18:45 -04:00

2.3 KiB

Translation status of the phrase page sets

Generated by build.py --reports; do not edit. Each text was drafted, then back-translated by a second reader who never saw the English, and the two compared. Only texts both agreed on are used ("agreed by two model readers; native review pending"); the rest stay in English, marked as such in every package.

Language Agreed Disputed Not yet translated Speaker-gendered
Spanish (es) 1075 / 1075 0 0 29
French (fr) 1075 / 1075 0 0 20
German (de) 1075 / 1075 0 0 0
Portuguese (Brazil) (pt) 1075 / 1075 0 0 30
Italian (it) 1075 / 1075 0 0 24
Chinese (Simplified) (zh) 1075 / 1075 0 0 0
Japanese (ja) 1074 / 1075 1 0 0
Arabic (ar) 1074 / 1075 1 0 0
Hindi (hi) 1075 / 1075 0 0 14
Russian (ru) 1075 / 1075 0 0 15

Pending (English only for now): nl, ko, he, ur, fa, bn, ta, te, ml, mr, kn, gu, pa, th, vi, id, ms, tl, pl, uk, ro, cs, sv, da, nb, fi, sw, am, ha, yo, zu, tr, el, hu.

Japanese: disputed, shown in English

  • 'OK': 'I'm OK / it's fine', not the lukewarm 'so-so' of まあまあ. / Natural, polite (です). It has a common second sense as a polite refusal, "No thank you, I'm good", so a partner offering something may hear "no". No gender marking. / In reply to an offer, 大丈夫です is commonly heard as a polite refusal ('no thanks, I'm fine'), the opposite of agreeing 'OK'; a partner may hear 'no'.

Arabic: disputed, shown in English

  • 'I did that for … years.': 'The number of years I spent on that: …' — general (job, hobby or activity), no noun after the number, no gender marking. / Stiff, form-like noun phrase rather than something a person says; it trails off with an ellipsis. Without vowel marks, "قضيتها" can be read as "I spent" (qaḍaytuhā), "you (m.) spent" or "you (f.) spent", so the speech engine may voice it as "you spent". "ذلك" is masculine "that", which constrains the next word to a masculine noun. Reads as the start of a statement, not a question. MSA register. / Not a first-person statement: it is the stock questionnaire phrasing 'number of years you spent on that', and unvocalized قضيتها is readily read or voiced as 'you spent'.