Plan Import Text

Plan: import any text, and make cards from its words

Source of truth: this wiki page.

Written 2026-10-05.

What the owner asked

Import any text (LingQ): paste text, tap unknown words, create cards with sentence context. think about this feature as well

The ROADMAP already lists "Sentence mining from imported text" under "Later".

What exists

• Decks can be added (#22), as files, and are kept by FileDeckStore under added/. A deck of mined words can live there.
• Readings and IPA are not made in the app. tools/transcribe.py writes them when decks are made, and the app only compares typed answers with them (romanised.dart). A pasted word has no reading until one is made.
• Meanings come only from the decks. The app has no dictionary.
• Words are split at spaces (wordsOf). That does not work for Thai, Chinese or Japanese (language-paths.md).

The feature

1. Paste or share text into the app, for a language being learnt.
2. The text is shown with each word marked:
   • known: a card the learner has learnt;
   • learning: a card being learnt;
   • in the course, not yet taught;
   • new: in no deck.
3. Tap a word to see its card, if it has one, or to make one:
   • the meaning: from the decks where the word is in one; otherwise typed by the learner, or from an offline dictionary (see below);
   • the sentence it was in, kept as the card's example, with its source;
   • the reading and IPA, if the app can make them; otherwise typed.
4. The cards go to a deck of the learner's own per language, such as "My words", and are learnt and reviewed like any other.
5. The text is kept to read again, with its words marked as they are learnt.

What it takes

1. A reader screen: words marked by state, tapping, and the sentence around a word.
2. Matching words to cards, forgiving inflection where it can: a form a grammar table lists matches its lemma.
3. A card maker that writes cards into the learner's own deck, with ids that stay unique and permanent (AGENTS.md).
4. Readings and IPA in the app: a Dart port of transcribe.py's rules, for the languages it knows (in-app-readings.md, which measures how accurate they are). Otherwise the learner types the reading.
5. Meanings, one of:
   • the learner types them (works offline, no licence question);
   • an offline dictionary per language, downloaded like the decks, built from Wiktionary through kaikki.org. That is CC BY-SA, the licence question settled for the decks in ADR-0040.
6. Tests for word states, matching, the card maker, and ids.

To decide

• Where meanings come from: typed, a downloaded dictionary, or both.
• Whether to port the reading and IPA rules to Dart, or ask the learner.
• Whether imported texts can be shared, or stay on the phone.
• The languages without spaces: wait for their word splitting, or leave them out.

Estimate

Part	Hours
Reader, word states, tapping, card maker	8–12
Readings and IPA in Dart, for the seven Indian languages	8–12
An offline dictionary from Wiktionary, per language	4–6 each

Confidence: low; most of it depends on the choices above.

See also: Roadmap · Plans-Index · Home