Resources & Data Sources

The Japanese-Multilingual Dictionary is the successor to the original EDICT file and contains well over 200,000 entries. Each entry lists the kanji and kana forms of a word, its part of speech, English glosses grouped into senses, usage notes such as "usually written in kana", field and dialect tags, and the frequency markers drawn from newspaper, loanword, and word list surveys.

The Japanese proper names dictionary, originally ENAMDICT, holds around 750,000 entries for surnames, given names, place names, companies, products, stations, and works of fiction, together with their readings and a classification of the name type.

A dictionary of more than 13,000 kanji covering the JIS X 0208, 0212, and 0213 character sets. For each character it records on and kun readings, readings used only in names (nanori), English meanings, stroke count, the classical radical, the school grade at which it is taught, a newspaper frequency rank for the 2,500 most used characters, the former JLPT level, variant forms, and index numbers in the major printed kanji dictionaries.

A decomposition of over 6,000 kanji into their visible component parts. Unlike the single classical radical listed in KANJIDIC2, KRADFILE lists every element you can see in a character, which is what makes it possible to find a kanji by picking the pieces you recognise.

KanjiVG (Kanji Vector Graphics) describes more than 11,000 kanji and kana as SVG drawings in which every stroke is a separate path, numbered in the order it is written and annotated with the component it belongs to. Ulrich Apel began the project in 2009 and it has been refined by volunteers ever since. Because the strokes are vectors rather than pictures, they can be animated, coloured, scaled, and traced without any loss of quality.

MeCab ("Yet Another Part-of-Speech and Morphological Analyzer") is the standard open source segmenter for Japanese. Given a sentence it returns each morpheme with its part of speech, conjugation type, dictionary form, and reading, choosing the most probable segmentation with a conditional random field model trained on annotated corpora.

IPADIC is the morphological dictionary originally built for the ChaSen analyser at the Nara Institute of Science and Technology with support from the Information-technology Promotion Agency. mecab-ipadic-NEologd extends it with hundreds of thousands of neologisms, proper nouns, product names, and slang harvested from the web, so recent vocabulary is segmented as a single word instead of being broken into fragments.

UniDic is a linguistically principled dictionary for MeCab developed at NINJAL, Japan's national language research institute. It is built around the "short unit word" and carries consistent lemmas, pronunciations, accent information, and orthographic variants for every entry. The contemporary written Japanese edition (unidic-cwj) is trained on the Balanced Corpus of Contemporary Written Japanese.

The vocabulary lists for levels N5 to N1 were compiled by Jonathan Waller for his JLPT Resources site from the vocabulary sections of past exams, and are the same lists many dictionaries use for their JLPT tags. The copy used here comes from the yomitan-jlpt-vocab project, which matched every word to its JMdict entry and normalised rare spellings to the common JMdict form. There are no official JLPT vocabulary lists, so these levels are an informed estimate rather than a guarantee of what appears on the test.

The grammar points for levels N5 to N1, their meanings, formation rules by word class, and example sentences were written and organised by our team from widely taught textbook grammar. The Japanese-Language Proficiency Test does not publish an official grammar list, so level assignments follow common study guides and may differ from other references. JLPT is a trademark of the Japan Foundation and Japan Educational Exchanges and Services, with which this site is not affiliated.

The number readings, the one-to-ten readings of the counters, the groups they are filed under, the colour values behind the colour cards, and the word-list collections (body, clock, connectors, demonstratives, family, feelings, frequency, house, position, questions, time, and weather) were assembled by our team from the forms taught in textbooks. The boxes above those lists, from the head down through the family, the house, and the other sets, were drawn up the same way. The words themselves are the JMdict entries. Traditional colour names do not have one official sample, so each dot is a close match.

The hiragana and katakana tables, including voiced, semi-voiced, combination, and extended katakana sounds, and the conversion between kana and Hepburn romaji used by the search box, were assembled by our team. Stroke order for each kana comes from KanjiVG.

The 214 classical radicals and their stroke counts follow the system established by the Kangxi Dictionary, compiled under the Kangxi Emperor and still the standard ordering for kanji dictionaries. KANJIDIC2 assigns each kanji its classical radical using this numbering.