Where does the data come from?
Acknowledgment
This application uses dictionary files from the Electronic Dictionary Research and Development Group (KANJIDIC2, JMdict and KRADFILE), used in conformance with the Group's license. The files are made available under the Creative Commons Attribution-ShareAlike 4.0 International license, and the data shown here is derived from them. Example sentences come from the Tatoeba Project (tatoeba.org) and are used under the Creative Commons Attribution 2.0 France license; they were written by Tatoeba's contributors, not by this application. The stroke-order diagrams are drawn from KanjiVG (© Ulrich Apel and contributors), used under the Creative Commons Attribution-ShareAlike 3.0 license. Word-track categories, teaching priority, reading preference and within-definition reading order are derived from the Corpus of Everyday Japanese Conversation (CEJC) short-unit vocabulary tables, version 2022.09, created by the National Institute for Japanese Language and Linguistics. Raw CEJC data is not redistributed. The JLPT vocabulary lists at tanos.co.uk (Jonathan Waller, CC BY), open-anki-jlpt-decks (MIT), and the OpenSubtitles 2018 frequency list from hermitdave/FrequencyWords (CC BY-SA 4.0) provide fallback ordering only for words CEJC does not observe. The glyph-origin notes, covering what each component of a kanji contributes and how the parts make the whole, are derived from English Wiktionary (en.wiktionary.org), used under the Creative Commons Attribution-ShareAlike 4.0 International license and written by Wiktionary's contributors. The Japanese names of the classical radicals and their positional variant forms are derived from Kanji Alive (github.com/kanjialive), used under the Creative Commons Attribution 4.0 International license.
The files
- KANJIDIC2
Kanji meanings, on/kun readings, stroke counts, school grades, and newspaper frequency ranks.
Electronic Dictionary Research and Development Group · CC BY-SA 4.0
- JMdict
Everyday vocabulary: written forms, readings, English meanings, and part-of-speech tags.
Electronic Dictionary Research and Development Group · CC BY-SA 4.0
- KRADFILE
The radical decomposition each kanji is built from, which drives parts-first ordering.
Electronic Dictionary Research and Development Group · CC BY-SA 4.0
- Tatoeba
Example sentences and their human English translations, used to show each grammar pattern in a real sentence.
The Tatoeba Project and its contributors · CC BY 2.0 FR
- KanjiVG
The stroke-order data behind every “how it’s written” diagram: each stroke as a path, in the order it is drawn.
© Ulrich Apel and contributors · CC BY-SA 3.0
- English Wiktionary
Each kanji's glyph origin: what each component contributes (meaning, sound, or shape), the sense that applies, and the plain-language explanation of how the parts make the whole. This covers both the crawled set (kanji-etymology.json) and a hand-authored layer (kanji-etymology-manual.json) that re-maps the same Wiktionary glyph-origin prose onto the visible modern glyph where the automated join could not, e.g., 時's phonetic, historically 之, now written 寺. Every hand-authored entry cites its source and stays CC BY-SA 4.0.
Wiktionary contributors · CC BY-SA 4.0
- Kanji Alive (japanese-radicals.csv)
Each Kangxi radical's Japanese bushu name (kana and romaji) and its positional variant forms.
Kanji Alive project · CC BY 4.0
- JLPT vocabulary lists (tanos.co.uk)
One of two proficiency lists used only in the fallback ordering for words absent from CEJC.
Jonathan Waller (tanos.co.uk) · CC BY
- open-anki-jlpt-decks
The second proficiency list used in the CEJC-unobserved fallback ordering.
jamsinclair and contributors · MIT
- FrequencyWords (OpenSubtitles 2018, Japanese)
A secondary frequency source used to order Library words that CEJC does not observe.
Hermit Dave (hermitdave/FrequencyWords) · CC BY-SA 4.0
- Corpus of Everyday Japanese Conversation (CEJC)
Observed lexical and pronunciation frequencies from 200 hours of everyday conversation. CEJC part of speech separates core vocabulary, conversational essentials, grammar and fillers; its counts order the word curriculum and interchangeable readings within each dictionary meaning.
National Institute for Japanese Language and Linguistics · Free for research and education; raw redistribution prohibited
Share-alike
CC BY-SA is share-alike. The dictionary files this app reads are adaptations of EDRDG's, so they have the same license. The app's own code reads that data without deriving from it, so the code itself is MIT. Tatoeba's sentences are attribution-only, with no share-alike.
KanjiVG's stroke data is CC BY-SA 3.0, a version behind the rest. A 3.0 share-alike work may be used in a 4.0 one, so the diagrams can be used alongside the dictionary data.
CEJC's raw conversation-frequency files are not included. The app ships only the reduced reading order used for its own vocabulary, and identifies that educational analysis separately rather than relicensing NINJAL's work as CC BY-SA.
Frequency comparisons never cross JMdict meanings. When every reading has the same sense coverage and CEJC provides at least 50 observations, a reading at or below 5% of usage moves to the Atlas's other dictionary readings rather than the teaching table.
Other places to learn
Saku does not teach everything. These are other people's sites and books, worth going to for what it leaves out.
Kana
- Tofugu: Learn hiragana
A mnemonic for every shape. It is free and takes about a weekend.
- Tofugu: Learn katakana
The same guide for the other set. Do it after hiragana.
Kanji & vocab
Grammar
- Bunpro
Grammar points drilled on a review schedule, ordered by JLPT level.
- Tae Kim's Guide to Japanese
Free and complete. It explains why the grammar works instead of only what to say.
Dictionaries
- Jisho
Look up a word, a kanji, or a stroke you can only draw.
Reading practice
- NHK News Web Easy
Real news, simplified, with furigana. Free, and new every day.
- Satori Reader
Graded stories with grammar notes in line. Best once kana is solid.
Textbooks
- Minna no Nihongo
The classroom standard. Buy the translation booklet with it; the textbook itself is all Japanese.
