Japanese Dictionary Tools: How to Look Up Any Word

Looking things up in Japanese is a skill, and it is not the same skill as knowing the word. A learner who can find any unknown word in fifteen seconds reads twice as much as one who gets stuck for five minutes and gives up — which means dictionary technique is one of the highest-return things a beginner can invest a weekend in.

There is also a specific problem that does not exist in French or Spanish: you cannot sound out an unknown character. 新聞 (shinbun, "newspaper") is a word you will meet on any news site, and nothing about the shape of those two characters tells you how it is pronounced. Until you can look a character up by shape, unknown words are invisible to you, no matter how many you have memorised.

The four kinds of tool

Most learners end up with two or three of these, for different situations.

Jisho.org is the one to set as your home page. It handles romaji, kana, kanji, and English input in the same box, indexes example sentences, and — the part that matters — has a proper kanji search with a radical picker and stroke-count filtering. Its weaknesses are that it needs a connection and that its example sentences are drawn from a mixed corpus of uneven quality.

Takoboto and Shirabe Jisho are the offline apps. Both let you draw a character with your finger and get candidates back, which on a phone is usually faster than any other method. Takoboto is Android and web; Shirabe Jisho is iOS. If you study on a commute with no signal, one of these is not optional.

The browser extension is the tool that changes how much you read. The modern choice is Yomitan, a maintained successor to the discontinued Yomichan; 10ten Japanese Reader (formerly Rikaichamp) is the lighter, simpler alternative, and Rikaikun is the older Chrome option that many people still have installed. All of them do the same thing: hover over a word on any web page and a popup shows the reading and the meaning, and often the dictionary form of a conjugated verb. For reading practice this is transformative — it removes the alt-tab to a dictionary tab that otherwise breaks your attention every ten seconds. It is also, honestly, a crutch: knowing every word is one hover away makes it easy to never consolidate any of them.

Why radical lookup and stroke-count lookup are different

This is the part that confuses people, and it is worth understanding because the two methods fail in opposite situations.

Kanji dictionaries are indexed by radical — the component that carries the character's rough meaning category. There are 214 of them in the traditional system, and each character has exactly one. 打つ (utsu, "to hit") contains the hand radical 扌 (te-hen), and so does 押す (osu, "to push") and 投げる (nageru, "to throw"). Look up the radical, and you get a list of every character built on it.

Within a radical group, characters are ordered by stroke count — specifically, by how many strokes remain after you exclude the radical. That is why the two methods are usually described as one procedure: radical first, then stroke count of the rest. It is the index of a physical dictionary, and it survived into the digital ones.

The limitation is that both methods require you to decompose the character: to see 扌 as a hand radical and to count strokes correctly. That is a skill, and it takes a few weeks of exposure — learning kanji is what builds it. If you cannot tell which part is the radical — which is common with characters that are mostly one dense block — radical lookup stalls.

Handwriting recognition fails the opposite way. It does not care about components or stroke count at all; it matches the shape you drew against a database of shapes. So:

• If you can picture the character but cannot decompose it, draw it. This is the phone case, and it is why drawing is the default method on mobile.
• If you can decompose it but cannot remember what it looks like, use radical plus stroke count. This is the desktop case.
• If you can neither, you probably have the character in front of you — in which case the fastest method is to copy it, or to type it with your IME.

A third system worth knowing about is SKIP (System of Kanji Indexing by Patterns), which classifies characters by their overall shape — left-right, top-bottom, enclosure, or solid — and is dramatically faster than radical-plus-strokes once you know it. Jisho supports it if you enable it. Most learners never bother, and that is a defensible choice.

The workflow for a kanji you met in the wild

You are reading something and hit a character you have never seen. Here is the order that wastes the least time.

1. Try the whole word first, not the character. This is the single most important habit. Kanji readings are context-dependent, so looking up 日 (hi, "day") on its own gives you a wall of possibilities; looking up 日曜日 (nichiyoubi, "Sunday") gives you one answer. If the text has furigana — the small kana printed above the kanji — read those and search the kana, which is always unambiguous.
2. If you are on a computer, type it. Switch your IME to Japanese, type the reading in romaji, and let the IME's candidate list find the character. For a word whose reading you know, this beats every lookup method. If you do not know the reading, you cannot do this — which is the whole reason the other methods exist.
3. If you are on a phone, draw it. Finger-draw the character into Takoboto or Shirabe Jisho. Do not worry about stroke order — recognition is shape-based, so a sloppy but correctly-shaped character usually works.
4. If drawing fails, go by radical. Pick out the component that looks most like a meaning-bearing part, find it in Jisho's radical list, then narrow by stroke count.
5. Confirm with the sentence, not the gloss. A dictionary entry is a list of possible meanings in rough frequency order. Your sentence decides which one is live. 切る (kiru) is "to cut" in most contexts, but in 電話を切る (denwa wo kiru, "to hang up the phone") the gloss you want is not "cut" at all. The example sentences are the part of the entry that actually teaches you the word.

What a dictionary will not tell you

Two gaps are worth knowing about because they cause real errors.

Pitch accent. Standard dictionaries do not show it, and pitch accent changes meaning: 箸 (hashi, "chopsticks"), 橋 (hashi, "bridge"), and 端 (hashi, "edge") differ mainly in the melody, not the sounds. When you need it, use OJAD, the Online Japanese Accent Dictionary from the University of Tokyo, which is free and shows the accent pattern for whole conjugated phrases, not just dictionary forms.

Transitivity. Japanese verbs come in pairs, and the dictionary will label them 他動詞 (tadoushi, "transitive") and 自動詞 (jidoushi, "intransitive") in Japanese terms you may not recognise at first. 開く has a transitive reading (akeru, "to open something") and an intransitive one (hiraku, "to open, to be open"). Getting the pair wrong produces sentences that are grammatical and mean the opposite of what you intended. Jisho marks the part of speech — learn to read those labels early.

Common mistakes for English speakers

• Searching a single kanji when you already have the word. The character is not the unit of meaning; the word is. Look up 新聞 (shinbun, "newspaper"), not 新 (shin, "new") on its own.
• Trusting romaji search too far. Romaji is ambiguous in ways that kana is not — long vowels are written several different ways, and the kana じ and ぢ both romanise to ji, as do ず and づ to zu. Search in kana when you can. The romaji input guide covers where the ambiguities actually bite.
• Using a machine translator as your dictionary. Pasting a whole sentence into a translator gives you a plausible English sentence and zero information about which word meant what, how it conjugates, or what part of speech it is. You learn nothing you can reuse, and you build the habit of reaching for it.
• Reading only the first gloss. The first sense is the most frequent one overall, not the one in front of you. Check the example sentences before you decide.
• Stopping to look up everything. Reading with a dictionary open on every word is transcription, not reading. Look up the words that block your understanding of the sentence and let the rest go — the ones that matter will come back.
• Looking a word up and not recording it. A lookup you never revisit is a lookup you will repeat. Save the word with the sentence you found it in; a bare word list is much harder to review than a sentence. Vocabulary memory covers what to do with the list afterwards.
• Never upgrading past the dictionary. The end state is recognising the word without looking it up. If you are still hovering over the same word in month six, it belongs in your review queue, not in your dictionary.

The setup worth having

For most English-speaking learners the whole toolkit is three things: Jisho.org bookmarked on the computer, a drawing-capable offline app on the phone, and a pop-up dictionary in the browser. That covers lookup, offline lookup, and reading in place. Everything else is preference.

What none of them do is make you remember anything. The dictionary's job is to remove the blocker; the remembering happens somewhere else, in review and in use.

Our sentence practice puts the reading and the meaning next to every word, so most lookups never happen — you type the sentence in romaji, find out immediately which word you missed, and keep moving. The romaji converter is the fallback for a reading you cannot work out, and the kana table is where the whole lookup skill starts, because you cannot search a word in kana until you can read kana. Once the lookups are fast, the next problem is retention — see vocabulary memory and reading practice.