Languages
Droid Sentence Miner is built for Japanese first, and it supports other languages as small modules. On first launch it asks What are you learning?: Japanese is preselected. Change it any time in Settings → Language.
What follows the language
Section titled “What follows the language”- Lookups use only that language’s dictionaries. Dictionaries of the other language stay installed but are hidden and unused, so you can switch back without reinstalling.
- Settings → Dictionaries shows that language’s dictionaries and one-tap recommendations.
- OCR clean-up keeps the language’s spacing and punctuation (English keeps spaces and
!?, Japanese uses full-width punctuation and joins wrapped lines without spaces). - Downloads: nothing for a language you don’t study. The APK carries only a few KB of code per language.
English
Section titled “English”- Dictionaries: install Wiktionary English (English → English, about 110 MB download) and Wiktionary English IPA (pronunciations, 3 MB) under Settings → Dictionaries. Small English → Portuguese, Spanish, German, French, Italian, Russian, Chinese, Japanese and Korean glossaries are there too.
- Words and phrases: tap a word and the longest dictionary match wins, up to five words, so looked up, ice cream and give up on open as phrases. Inflections are undone with parts of speech: running → run, happiest → happy, wolves → wolf, looked up → look up.
- Pronunciation: the IPA is shown under the headword and goes into the card’s IPA field (or Reading, on a note type without one). Word audio comes from Wiktionary’s recordings on Wikimedia Commons (US first).
- OCR: without downloads, English is read by the built-in ML Kit model. For games and subtitles install the English OCR models (23 MB) in Settings → Language: PaddleOCR made about a third of ML Kit’s word errors in our test (3.5 % vs 12.2 %) and is faster.
German
Section titled “German”- Dictionaries: install Wiktionary German → English (17 MB) and Wiktionary German IPA (1.5 MB). Wiktionary German (German → German, 55 MB) and German → Portuguese, Spanish, French, Italian, Russian, Chinese, Japanese and Korean glossaries are there too. Which is which: dictionary list.
- Words and phrases: whole words and phrases of up to four words, hyphenated compounds (E-Mail). Nouns keep their capital; a capitalized word at the start of a sentence and ALL-CAPS game text are looked up too. Wiktionary lists most inflected forms itself (ging → gehen, Häuser → Haus); regular ones are undone on top: gemacht → machen, aufgemacht → aufmachen, anzufangen → anfangen, schnelleren → schnell, Lehrerinnen → Lehrerin. Separated prefixes (Ich stehe früh auf) aren’t joined yet.
- Cards: the Gender field gets the noun’s article (der Hund, die Katze), the IPA field the pronunciation. Word audio: Wiktionary’s recordings (standard German first, then Austrian and Swiss), then Lingua Libre.
- OCR: the built-in ML Kit model reads umlauts and ß; the optional OCR models (23 MB, the same PaddleOCR pack as English) work too. Hyphenated line breaks (Wör- / terbuch) are joined.
Japanese
Section titled “Japanese”Everything else on this site describes Japanese: game-trained meikiocr, pitch accent, Jitendex, JapanesePod101 word audio, vertical book pages.
More languages
Section titled “More languages”A language is a small Kotlin module plus a list of recommended dictionaries and, optionally, a downloadable OCR pack. If you want to bring your language, see the contributor guide.