Skip to content
Droid Sentence MinerDroid Sentence Miner

Anki mining

Droid Sentence Miner builds the whole card in one press. It supports two backends and fills existing note types by field name, the way Yomitan and GameSentenceMiner do.

Backends

Section titled “Backends”

Cards, images and audio go through the official AnkiDroid API. No network is needed.

  1. Install AnkiDroid 2.16 or newer.
  2. In AnkiDroid open Settings → Advanced → AnkiDroid API and enable it.
  3. In Droid Sentence Miner open Settings → Anki, choose AnkiDroid, and grant the API permission.
  4. Pick a deck and note type.

The app asks AnkiDroid to create media files and the note. It retries with a built-in Basic note type if a just-created note type is refused by the backend; your card is still created.

Fields

Section titled “Fields”

The mapper uses aliases in preference order, not note-type field order. A card can contain:

Content Recognized field names (in preference order)
Word Expression, Word, Vocab, Vocabulary, Term, Target Word, Key
Reading Reading, Word Reading, Kana, Hiragana, Furigana
Glossary Meaning, Glossary, Definition(s), English, Translation, Explanation
Context sentence Context Sentence, Sentence, Context, Example, Example Sentence, Sentence JP, JP Sentence
Screenshot Image, Picture, Snapshot, Screenshot, Photo
Sentence audio Audio, Sentence Audio, Voice, Sound, Audio Clip
Word audio Word Audio, Expression Audio, Vocab Audio, Vocabulary Audio, Term Audio, Audio Word
Pitch graph (Japanese) Pitch Graph, Graph, Pitch Accent Graph(s), Pitch Accent, Pitch, Pitch Accents
IPA (English, German) IPA, Pronunciation, Phonetic(s), Transcription, then Reading, Word Reading
Gender (German) Gender, Article, Artikel, Genus — the noun’s article: der, die, das
Source Source, Origin, Title, URL

Field names are normalized (lowercase, punctuation removed), so Context Sentence and context_sentence match. Only fields that exist on the chosen note type are filled; the rest are skipped. Plain Basic note types with two fields use a compact Front/Back layout instead.

Per language

Section titled “Per language”

Deck, note type and tags are kept per study language — Japanese keeps the settings you already had. Each language has its own note type, created on the first card if it doesn’t exist: Droid Sentence Miner (Japanese: Reading and Pitch graph), Droid Sentence Miner (English) (IPA instead of Reading, no pitch graph) and Droid Sentence Miner (German) (Gender and IPA), in the decks Droid Sentence Miner, Droid Sentence Miner::English and Droid Sentence Miner::German.

Leaving contents out

Section titled “Leaving contents out”

In Settings → Anki → Fields, set a field to Nothing and that content is left out: it isn’t even fetched (no word-audio download, no sentence clip, no screenshot). For simple cards, map only the word, the sentence and the glossary. The word is always added.

If the automatic mapping is wrong, open Settings → Anki → Fields, load the note type, and pick what each field receives. A custom map replaces name matching for that note type and is saved per model; Reset to automatic returns to the aliases above. The Picture max height setting (default 462 px, 0 = original) scales the screenshot with its aspect ratio — it never upscales.

Anime Cards

Section titled “Anime Cards”

Anime Cards is a common Yomitan/GameSentenceMiner note type: Word, Graph, Reading, Glossary, Context Sentence, Word Audio, Audio, Picture, Sentence.

  • The sentence goes to Context Sentence, not Sentence. Anime Cards shows {{Sentence}} on the front when it is filled, which would turn a word card into a sentence card. The app keeps Sentence empty on purpose, matching notes made by Yomitan/GSM.
  • Audio is used for word audio when both Audio and a sentence-audio field exist.
  • If the note type has no word-audio field, the [sound:] tag is appended to Reading.

Foreign note types get the sentence as plain text instead of the Yomitan glossary wrapper.

Edit before adding

Section titled “Edit before adding”

The ✎ joined to + Anki (on the popup header and on each headword’s button) opens the card editor before anything is added:

  • Sentence — edit the text (the mined word is bold), or tap one of the chips under it to keep only that sentence of a multi-sentence line. Reset brings the original back.
  • Sentence audio — the line’s clip as a waveform, with the handles already on the detected voice (the rest is usually room noise or music you sat through while reading). Play it, move the handles, or switch the audio off. A silent clip (an unvoiced line) starts switched off.
  • Add to Anki mines the edited card; the popup’s button turns Added as usual.

Over a game or video the editor is its own window and the keyboard opens there; Back closes it.

Which dictionaries go on the card

Section titled “Which dictionaries go on the card”

By default the card gets the definitions of the dictionary at the top of Settings → Dictionaries. In the popup, that dictionary’s name shows ✓ On card and the others show Use on card. Tap another dictionary’s name to put only its definitions on this word’s card; tap it again to go back to the default.

Settings → Cards → Glossary sets the default: only the first dictionary (default), all dictionaries without their names, or all under their names.

Duplicate check

Section titled “Duplicate check”

Before adding, the popup checks the configured note type for a note whose first field matches the word. If one exists, the button shows ✓ In Anki and no duplicate is created. This uses the same rule addNote refuses on, including across AnkiDroid and AnkiConnect.

Mining keeps running if you close the popup immediately after tapping Anki. The card is created on the app’s scope, not the popup’s, so it is not cancelled by the popup closing.

Word audio

Section titled “Word audio”

The word’s pronunciation is fetched from the same sources Yomitan uses for Japanese, in order:

  1. JapanesePod101 direct audio (audiomp3.php); kana-only words query by kana. The silent placeholder clip is rejected by SHA-256.
  2. JapanesePod101 dictionary search. Results are split on the dc-result-row class.
  3. Jisho audio_{term}:{reading}.
  4. An optional custom URL tried first, with {term} and {reading} placeholders.

A custom URL is useful for a local audio server or a paid source. If nothing is found, the card is still created without the word clip.

Pitch accent graph

Section titled “Pitch accent graph”

The pitch graph is the same inline SVG Yomitan writes with {pitch-accent-graphs}. It is filled into a pitch/graph field such as Anime Cards’ Graph. It reads pitch data from any installed Yomitan pitch dictionary; Kanjium’s accents are a one-tap install. If no pitch data exists for the word, the graph is skipped.

Mining history

Section titled “Mining history”

The Home tab shows recent cards and whether each one reached Anki. Use it to spot a backend that is offline or a card that was refused.

Section titled “Related”