Kanji dictionaries for learners: paper, app or both?

Sooner or later every learner meets a character they cannot read and cannot type. It might be in a novel, on a form, or in a draft someone has asked you to check. You cannot paste what you cannot type, and you cannot look up a reading you do not know. That is the problem a kanji dictionary solves: it lets you start from the shape of a character and arrive at its readings, its meanings and the words it appears in.

Learners have more ways to do this than ever: thick paper dictionaries arranged by radical, learner's dictionaries written in English, phone apps that let you draw the character with a finger, and camera lookup that reads it straight off a page. Each method has a weak spot. This piece goes through them, then suggests a setup that covers most needs without crowding the desk.

Before you start

1. RkanjiRemembering the Kanji 1A keyword and story method for learning to recognise and write the common kanji.Find it at Books-A-Million

Two kinds of dictionary, and a third for learners

Japanese has separate dictionaries for words and for characters. A 国語辞典 (kokugo jiten) is a monolingual word dictionary: you look a word up by its reading, in kana order, and it gives meanings and usage. A 漢和辞典 (kanwa jiten) is a character dictionary: each entry is a single kanji with its on and kun readings, its meanings, its radical and stroke count, and a list of compounds that use it. Japanese students use both at school, and the kanwa jiten is the one built for the "I cannot read this" moment.

Learner's kanji dictionaries in English sit between the two. They are organized for people who do not yet think in radicals, the entries give English meanings, and the example compounds are chosen for usefulness rather than completeness. For most readers of this site, one of these, or an app that does the same job, is the right first tool. A full kanwa jiten earns its shelf space later, when you start to care about rare characters, variant forms and the history behind a meaning.

Three ways to find a character by its shape

By radical

Traditional character dictionaries sort kanji under 214 radicals (部首, bushu), a system inherited from the Chinese 康熙字典 (the Kangxi dictionary). The radical is the component the dictionary uses as the character's heading: 氵 (sanzui, the water radical) in 海, 泳 and 池, or 木 (ki, "tree") in 林, 板 and 校. Under each radical, characters are listed by the number of remaining strokes.

The method is reliable once you can spot the radical, and that is the catch. The radical is not always the most obvious part. 問 (mon, "question") is filed under 口 (mouth), not under 門 (gate), while 間 (aida, "interval") is filed under 門. 聞 (kiku, "hear") goes under 耳 (ear). Learners spend a lot of time flipping between plausible headings. A good dictionary softens this with a radical chart inside the cover, cross-references, and notes on where awkward characters are filed.

By stroke count

Every character has a total stroke count (総画数, sōkakusū), and dictionaries keep an index of characters sorted by that number. When the radical will not reveal itself, you count the strokes and scan the list.

This works only if you count the way the dictionary does, and learners often miscount. 口 is three strokes, not four, because the top and right side are a single stroke. 阝 is three strokes, 糸 is six and 女 is three. If you know standard stroke order, counting is quick. If you do not, a stroke-count index will eat your time, which is one more reason to learn stroke order early.

By overall pattern

Some learner's dictionaries and apps offer lookup by overall shape. The best known is SKIP-style lookup (System of Kanji Indexing by Patterns). You first decide whether a character splits into left and right parts, top and bottom parts, an enclosure and what is inside it, or none of these, and then count the strokes in each part. 校 is a left and right character with 4 strokes on the left (木) and 6 on the right (交). You never need to know which part is the radical, which is why beginners find the method friendly. Its weakness is borderline shapes, where two people may divide the same character differently; good implementations accept more than one code for those.

Paper: what it still does well

A paper dictionary is slow, and the slowness is part of its value. To find a character you have to look at it properly: find the radical, count the strokes, compare it with its neighbors on the page. People who look things up on paper tend to notice the parts of characters, and that pays off later, when a new kanji turns up and you can guess at its meaning or reading from its components.

Paper also shows you things you did not ask for. The compounds listed under 生 or 行 run to whole columns, and scanning them builds vocabulary in a way a single search result does not. It has no notifications, no battery and no screen glare, which matters more than it sounds during a long reading session.

The drawbacks are real: weight, time, and the fact that a printed book cannot read a photo or recognize a scribble. For the moment when you need an answer in ten seconds, paper loses.

Apps: handwriting, components and the camera

Dictionary apps mostly offer four routes in:

• Handwriting recognition. You draw the character with a finger or a stylus and the app suggests matches. For most people this is the fastest method. Some recognizers are strict about stroke order and stroke count and miss a character drawn out of order; others are more forgiving. All of them reward neat writing in standard order.
• Component lookup. You tap the parts you can see, such as 氵 and 毎, and the app lists every character that contains them. It works even when you cannot tell which part is the radical.
• Copy and paste. If the character is already text on a screen, paste it and you are done. Many apps also split a pasted sentence into words and look each one up.
• Camera lookup. The app reads characters from a photo. It is handy for menus, signs and printed pages, and less reliable on handwriting, small print or vertical text with ruby.

Apps usually show information a pocket paper dictionary leaves out: whether a character is on the Jōyō list (常用漢字表, 2,136 characters), the school grade in which it is taught, a stroke order animation, and links from the character to full word entries. For anyone checking someone else's text, that Jōyō tag is the useful one. If a draft for a general audience uses a character outside the list, it is worth asking whether kana or a ruby reading would serve the reader better.

The weak spot is data quality. Apps differ, and a free one may draw its readings and example words from sources of uneven reliability. Treat a strange result as a reason to check a second source, not as an answer.

Side by side

Where a kanji course book fits

A dictionary is for finding a character you do not know. A course book is for learning characters in an order that makes them stick. These are different jobs, but a good course book also works as a small reference, because its indexes let you find a character you have already met and reread its entry.

The Kodansha Kanji Learner's Course covers the full Jōyō set and some common characters beyond it, in an order built around components. Each entry gives a short mnemonic, the readings and a selection of common words. Because the book introduces parts before the characters that use them, it doubles as training for dictionary lookup: after a few hundred entries you start to see 氵, 言 and 糸 everywhere, and radical and component searches stop feeling like guesswork.

Remembering the Kanji 1 takes a different route. It teaches the meaning and the writing of each character, tied to an English keyword and a short imagined story built from the character's parts, and leaves the readings to a later volume. Learners who like it value how quickly they can recognize and write a large number of characters; those who do not would rather learn readings and words from the start. Either way, it trains the eye for components that every lookup method depends on.

Neither book replaces a dictionary. Both make the dictionary faster to use.

A setup that works

For most learners from intermediate level up, a sensible desk looks like this:

1. One app with handwriting and component lookup, for the ten-second questions that come up while reading or checking text.
2. One course book, worked through steadily, so that the shapes stop being strangers.
3. A paper dictionary, either a learner's dictionary or a full kanwa jiten, for the sessions when you want to look at a character properly, see its family and read down its compounds.

If you check Japanese text rather than study it, add one habit. When you look up a character in a draft, note whether it is on the Jōyō list and whether its reading is obvious to a general reader. Those notes feed straight into decisions about kana and ruby, and into the style sheet for the document. For the grammar side of the same desk, see the grammar dictionaries worth keeping close; the full set of desk books is on the writing desk page.

Questions readers ask

Do I need a paper dictionary at all?

No. Plenty of learners reach a high level with apps alone. The case for paper is not that it finds characters better, but that it makes you look harder at them. If you already study character structure through a course book, an app may be all you need.

Why does handwriting recognition keep missing my characters?

Usually because of stroke order or stroke count: an extra stroke, two strokes run together, or a component written in an unusual order. Write a little more slowly and in standard order. If one character keeps failing, switch to component lookup.

Should I learn all 214 radicals by heart?

Not as a list. Learn the common ones as you meet them inside characters, and the rest will follow. Knowing that 扌 is "hand" and 忄 is "heart" helps with meaning far more than knowing a radical's number.

Is SKIP-style lookup the same as radical lookup?

No. Radical lookup needs you to identify the component the dictionary uses as a heading. A SKIP-style code describes the overall shape and the stroke counts of its parts, so it works even when the radical is unclear.