Dictionary & Kanji

The Dictionary & kanji view is a single search box over the active language's dictionary. For Japanese it answers with both JMdict word entries and kanji reference cards, including animated stroke-order diagrams; in a pack language it searches the pack's dictionary. It is also where you can add a word to spaced repetition without having met it in a book.

Searching

For Japanese, type kanji, kana, or rōmaji — 猫, ねこ, and neko all reach the same entry (rōmaji is transliterated as you type, so tabemashita works too). A conjugated query like 食べました resolves to its dictionary root, and a banner above the results explains the form: the typed form, its root, what kind of word it is, and the grammar of its tail. Results update as you type; the ✕ button clears the query and refocuses the box.

In a pack language the lookup is folded — case and accents are ignored — and an inflected query resolves through the pack's grammar table to every candidate lemma: suis surfaces both être and suivre. Forms the table doesn't list fall back to suffix rules the builder learned from its own data.

Koine Greek additionally accepts betacode/Greeklish in the box: logos finds λόγος. Betacode letter values (h = η, q = θ, w = ω, x = χ, c = ξ, y = ψ, f = φ) and the common digraphs (th, ph, ch, ps) both work, and betacode diacritic marks are accepted and dropped — matching is accent-insensitive either way.

Results build in tiers of match closeness: first the word the query is a form of, then exact matches, then prefix matches (so ねこ still surfaces ねこじた and friends; the prefix tier is capped at 30 entries). Corpus frequency orders the words within each tier — the everyday être beats the archaic estre — and a prefix match never outranks an exact one.

If the reference data was never downloaded (you chose "Continue without dictionary" at first run), the view shows a no-dictionary notice instead of results; retry the download from the banner. See Getting-Started.

Word entries

The left column lists word entries. Each shows:

• the headword, with its kana reading in parentheses when they differ,
• a status chip (unknown / learning / known / ignored) if you have already met the word while reading,
• the first three senses, with their English glosses.

Packs built from Wiktionary carry IPA pronunciation: with Settings → Reading → "Show IPA with dictionary entries" enabled (off by default), it appears where the kana reading would for Japanese.

Learn (SRS) from search

Each entry carries a ➕ Learn (SRS) button (hidden when the word is already in learning status). Clicking it:

1. Runs the headword through the morphological analyzer to derive the canonical word key (lemma, reading, part of speech) — the same key the reader uses, so the search hit and the word you later meet in a book are one record, not duplicates. If analysis yields nothing, the headword itself is used as a noun.
2. Creates the word record if it has never been seen.
3. Starts a review card without a sentence context. Cards created while reading show the sentence the word came from; cards created from search have none until you meet the word in a book. See Reviews-and-SRS.

Kanji cards

Character cards exist only for Japanese: in a pack language no cards appear, and the word list takes the full width of the view.

For Japanese, the right column shows up to six kanji cards. They are chosen from the kanji characters in your query, in order; if the query contains no kanji (a kana search), the cards come from the headwords of the top three word hits instead — so searching ねこ still produces the card for 猫.

Each card contains:

Grade, JLPT, frequency, and variants only appear when KANJIDIC2 records them for that character.

Stroke-order diagrams

When KanjiVG has data for the character, the card animates the stroke order: the character draws itself one stroke at a time — the stroke being traced highlighted in accent blue with a pen tip, the not-yet-drawn strokes ghosted so the glyph keeps its shape — and loops when finished. Scrolling over the character scrubs it stroke by stroke (one wheel notch is one stroke; scrolling down advances) and briefly pauses the auto-play while you scrub.

KanjiVG covers roughly half of KANJIDIC2's characters. For kanji it does not cover — mostly rare and archaic ones — the card shows the character large instead of a diagram; everything else on the card is unaffected.

Kanji chips in the reader

In the reader's word panel, each kanji of the selected headword appears as a chip under the headword; clicking a chip leaves the book and opens this view with that character as the query, card and animation included. Leaving the reader this way credits the partially read page like a pause. See Reading.

Data sources and licenses

The kanji data is downloaded once at first run into the app's data directory, alongside JMdict, and imported into the local database; after that, lookups are fully offline.

KANJIDIC2 is fetched from EDRDG's canonical URL (the file is regenerated daily; no pinned version exists). KanjiVG is pinned to an immutable GitHub release tag. Word entries come from JMdict via the jmdict-simplified project — see Getting-Started for the full reference-data list.

Pack dictionaries ship their own data and licenses: the Koine Greek pack is built from MorphGNT, and packs built from Wiktionary use kaikki.org's Wiktextract data (CC BY-SA 4.0 & GFDL) and hermitdave's FrequencyWords lists (CC BY-SA 4.0). Every attribution — plus each installed pack's license line — is consolidated in Settings → General → About.