How to Look Up a Kanji You Cannot Read

Why an unknown character stops you

Reading Japanese is mostly smooth until one character breaks the line, and then everything after it becomes guesswork. The frustrating part is not the character itself but the missing route in: an alphabet lets you sound out a word you have never seen, while a kanji gives you nothing to type into a search box. That gap makes learners either skip the word or abandon the text.

The fix is procedural rather than clever. There are five practical ways into an unknown character, and none of them requires you to already know the reading. Once you have used each one a few times, you stop deciding which to try and simply reach for the one that fits where the character happens to be sitting.

Match the situation to the lookup route

Most learners settle on one route and force it everywhere, which is why lookups feel slow. Handwriting input on a blurry printed sign is painful, and camera recognition on a handwritten note usually returns a similar-looking character rather than the right one. Two minutes spent learning the routes you currently avoid saves far more time than getting faster at the one you already use.

Handwriting input when nothing can be selected

Every phone and computer can add a Japanese handwriting keyboard, and it is the most forgiving route because it accepts imperfect drawing. You trace the character roughly and pick from the candidates offered. It tolerates wrong stroke order surprisingly well, though correct order does improve the suggestions, and it copes with characters you can see clearly but cannot name.

The skill worth building is reading the candidate list rather than the drawing itself. The system usually offers several visually similar characters, so compare the small details before tapping: an extra horizontal stroke, a dot where you drew a line, a component on the left rather than underneath. Choosing the wrong candidate is the most common reason a lookup returns a word that makes no sense.

• Add the Japanese handwriting keyboard once, before you need it, not during a lookup.
• Draw large and slowly; cramped drawing produces worse candidates than untidy drawing.
• Write the components in a sensible order: left before right, top before bottom.
• Compare candidates against the original for stroke count and component placement before tapping.
• If nothing matches, redraw the character with one component simplified rather than starting over.

Counting strokes without guessing

Stroke count is the backbone of paper lookups and a useful filter in software. The rule is simple to state and easy to get wrong: one stroke is one contact between pen and paper, from the moment it lands to the moment it lifts. Many shapes that look like separate lines are written without lifting, which is why beginners consistently count too many.

Treat your count as a range rather than a fact. If you count eleven, search ten to twelve, and scan the results by shape. That habit removes almost all the frustration of paper dictionaries, where an off-by-one count sends you to the wrong page and makes the method feel unreliable when it is only imprecise.

• A line that turns a corner and continues is usually one stroke, not two.
• A vertical line ending in a hook is one stroke, hook included.
• Count the enclosing box of a character like the one for country as three strokes, not four.
• Count familiar components as chunks you already know rather than line by line.
• Search a band of two either side of your count, and compare shapes rather than trusting the number.

Finding the radical of an unfamiliar character

The radical is the component a dictionary files the character under, and in most characters it is the part that carries a broad meaning while another part hints at sound. You do not need to know the official radical to use radical search in software, which lets you tick every component you can recognise and narrows the results as you go. That is a far lower bar than identifying the single correct radical.

Position is the useful clue. Radicals commonly sit on the left, along the top, underneath, or wrapped around the outside, and a handful of frequent ones cover a large share of everyday characters: the water radical of three drops, the hand radical, the person radical, the tree radical, the speech radical, the heart radical, the thread radical, and the metal radical. Learning to name these on sight makes both paper and digital lookups faster.

• Split the character into the largest parts you recognise before naming anything.
• Check the left edge first; that is where the meaning radical most often sits.
• In digital radical search, tick two or three components you are sure of rather than hunting for the correct one.
• If the character has an enclosing shape, the enclosure is usually the radical.
• Note the radical when you file the character in your own list; it makes the next lookup faster.

Camera, screen text, and copy and paste

Camera recognition is unbeatable for printed material in the real world: menus, station signs, forms, medicine packets. Hold the phone square to the page, get the whole word in frame rather than one character, and keep the light even. It fails predictably on handwriting, stylised logo type, low contrast, curved surfaces, and vertical text squeezed into a narrow column.

When the text is already on a screen, do not photograph it. Copying digital text gives you the exact characters, which removes recognition error completely, and most phones can now select text inside an image on screen too. If you can select it, select it; if you can only see it, recognise it; if it is on paper in front of you, use the camera. That order costs nothing and prevents most wrong lookups.

• Capture the whole word, since a lookup of one character out of context often returns the wrong entry.
• Photograph vertical text in short segments rather than one long strip.
• Paste copied text straight into a dictionary rather than a general search box for cleaner results.
• Save the sentence you found the word in, not just the word; you will need the context later.
• For a form or an official letter, photograph the whole page first so you can re-check terms afterwards.

Reading the entry you land on

The distinction that matters most is between the character's meaning and the word's meaning. A gloss such as bright, hand, or air describes the character in isolation, and the word you are reading may have drifted a long way from that. Read the example words before you decide you understand the sentence, because the entry is answering a slightly different question from the one you asked.

One character, several readings

One character can carry a striking number of readings, and this one is the classic example. Notice the pattern rather than the list: standing alone or followed by hiragana it takes native readings, while inside a compound it shifts to a borrowed reading that also changes shape between sei and shō depending on the word. You cannot deduce which applies from the character alone, so the reading always comes from the word.

This is why writing single characters on flashcards with a column of readings works poorly. Record the word you actually met, with its reading and the sentence around it, and let the character's other readings arrive with the next word that uses them.

The compound trap

The most common lookup failure is not a wrong character but a wrong unit. You meet two characters, look up each one, add the meanings together, and produce something that is nearly right and therefore hard to doubt. Meanwhile the compound has a settled reading and a settled meaning that the two entries never mentioned.

Some words go further and take a reading assigned to the whole word rather than to its parts, so no amount of per-character analysis will produce it. These are frequent, everyday words, not obscure ones, which is precisely why they trip up beginners.

Before you split a word, try the whole string. If the dictionary returns a clean entry with a single reading and example sentences, you are done. Split into single characters only when the whole word returns nothing, and treat the result as a hypothesis to confirm against the sentence rather than an answer.

Asking a person how it is read

Asking is not the failure route. It is the correct route for names, for handwriting, and for anything on an official form, because those are exactly the cases where software guesses badly. Japanese speakers ask each other how names are read constantly, so the question carries no embarrassment at all.

Point at the character while you ask, and be specific about what you want: a reading, a meaning, or the same word written in hiragana. A vague question gets a vague answer, whereas asking for the reading in hiragana gets you something you can write down and search later.

Expect a short answer rather than an explanation. To the first question you will usually hear the reading on its own, as in 「かわ」です, and to the name question something like 「たなか」と読みます. If the reply is too fast, the request to say it slowly works better than asking again, because repeating the question often produces the same speed a second time.

One caution worth knowing: asking a colleague how to read their own surname is completely normal, but asking a stranger to read a document over your shoulder is not. For anything private, photograph it and look it up yourself.

Saying what you found

Finishing a lookup silently wastes most of its value. Say the reading aloud, then say one sentence describing what you found, because producing the word once fixes it far better than reading the entry twice. These lines are also what you will genuinely say when someone asks whether you managed to work the character out.

Keep a short running list of characters you have looked up more than once, with the word you met and the sentence it came from. Repeat lookups are the clearest signal of what to learn next, and the list costs nothing to build because you are already doing the work each time.