Explanation
Where the data comes from
The four sources, their terms, and what you must credit.
Each source has its own terms
| Entry point | Made from | Terms |
|---|---|---|
grades | KANJIDIC2, of the Electronic Dictionary Research and Development Group | CC BY-SA 4.0 and the Group's conditions |
jlpt | Jonathan Waller's JLPT Resources, tanos.co.uk | CC BY |
kanji | KANJIDIC2, with Waller's levels | CC BY-SA 4.0 and the Group's conditions; the levels CC BY |
look-alike | Lars Yencken's kanji confusion datasets, and scores by shape made from KanjiVG's strokes | CC BY-SA 3.0 (Yencken's scores are CC BY 3.0) |
words-common, and the pieces words fetches | JMdict, of the Electronic Dictionary Research and Development Group | CC BY-SA 4.0 and the Group's conditions |
entries | JMdict, the whole entries | CC BY-SA 4.0 and the Group's conditions |
examples | The Tatoeba project's sentences, linked to JMdict's senses by the Group's JMdict_e_examp | CC BY 2.0 FR for the sentences, each credited to its author; CC BY-SA 4.0 for the links |
strokes | KanjiVG, by Ulrich Apel and its contributors | CC BY-SA 3.0, credited where the strokes are shown |
JMdict's glosses in other languages are not used: the Group's licence puts only the Japanese and English parts under CC BY-SA 4.0 and says the others are under their compilers' own copyright. The full notices are in NOTICE.md, and each file's address, date and SHA-256 in sources.json.
What you must do
A page that shows the data credits it. A page showing KANJIDIC2's or JMdict's data must say so on that screen, and a project that passes on grades, kanji or the words passes on the share-alike terms for those tables. Code that merely imports them is not changed by that. NOTICE.md has wording to use.
The words, and keeping them current
The words are JMdict's, which the EDRDG asks a package to keep current: the 1st of every month a workflow fetches the newest JMdict, writes the words again and, if they changed, leaves them on a branch for a patch release. The release date is in the data (release), and the date and SHA-256 of the file used are in data-sources/sources.json. A page that shows a word credits JMdict, and NOTICE.md has wording to use.
The JLPT levels are one person's
The test has published no list of kanji since its present five levels began in 2010. Waller's lists are the ones people use, and they are a reconstruction: a level here says which of his lists holds a kanji, not what the test will ask.
School grades and the numbers
KANJIDIC2 numbers grades 1 to 6 for elementary school, 8 for the rest of the jōyō kanji, 9 for kanji allowed in names, and 10 for name variants of jōyō kanji. Grade 10 includes compatibility characters that look the same as a common kanji and are a different code point: U+FA45 looks like 海, U+6D77. A text normalized to NFC holds the common one.