Japanese counter-word (josūshi) suggester & reading lookup

Who checks counter words and what they are working on

Counter-word lookups are not just for language students. Translators verifying Japanese output, editors checking copy written by non-native speakers, curriculum designers building JLPT study materials, and localisation engineers writing test strings all use this type of tool. In those professional contexts, the text being verified may include draft copy that is not yet public — a product description for an unreleased item, a chapter from an unpublished textbook, a localised marketing headline.

Even a single noun input (‘犬’, ‘鉛筆’, ‘会議室’) carries implicit context about what someone is writing. A string of animal names might mean a children’s book draft; a series of business nouns might mean internal documentation. The noun alone is not sensitive, but the habit of sending all verification queries to an external server is worth reconsidering.

The risk when counter-word services use external NLP APIs

Some counter-word tools route input through a server-side morphological analyser — MeCab, kuromoji, or Sudachi running on a remote machine — to identify the noun category before returning the appropriate counter. That means every noun you submit travels over the network, is processed on hardware you do not control, and appears in a log.

Free NLP APIs frequently aggregate usage data for ad targeting or model training. A service that receives noun queries can build a picture of what content categories are most common among its users — useful for monetisation, but not something users typically expect when they are just checking if a dog is counted with 匹 or 頭.

Static dictionary lookup with no network dependency

This tool bundles a noun-to-counter mapping (~200 entries, 30 counter types) as a static JSON object compiled into the page at build time. When you enter a noun, the browser compares it against the local dictionary and returns the matching counter along with the complete reading table — including all irregular rendaku (連濁) and sokuon (促音化) forms for numbers 1 through 10, 20, and 100. No external NLP API is called at any point.

The irregular reading table (e.g. 3本 → sanbon, 6本 → roppon, 10本 → juppon) is also embedded locally. Even if your noun is not in the dictionary, the fallback flow — picking a counter manually from the full list and looking up its reading — runs entirely in the browser. Open DevTools → Network while using the tool: no requests fire in response to noun input.

Practical workflow for editors and language learners

Bookmark this tool alongside your Japanese writing environment. When you hit a counter word you are unsure about, type the noun here, confirm the counter and its reading, then return to your draft. The round-trip takes seconds, and the noun you checked never leaves your browser.

For JLPT preparation, running through the counter table for common nouns (animals → 匹/頭/羽, thin flat things → 枚, long thin things → 本, machines → 台) gives you a compact drill that works offline after the first page load. The irregular-reading section — which is the part learners consistently find hardest — is where this kind of reference is most useful.

The phonology behind rendaku, sokuon, and handakuon shifts

Counter readings that change with the number (3 本 sanbon, 6 本 roppon, 1 本 ippon) are not arbitrary memorisation: they follow regular Japanese phonological rules. 3 + 本 shifts the /h/ to /b/ because the moraic nasal ん triggers rendaku (連濁, sequential voicing) on the following voiceless stop. 6 本 roppon shows sokuon-ka (促音化, gemination) where /k/ at the end of roku becomes a small tsu, with the following ho shifting to handakuon po. 1 本 ippon follows the same pattern: ichi + hon → ip + pon.

The shift is predictable from the counter’s initial consonant. h- counters (本 hon, 匹 hiki, 分 fun, 票 hyō) shift to rendaku b- and handakuon p-. s- counters (冊 satsu, 歳 sai, 才 sai) shift to z- under voicing. k- counters (個 ko, 回 kai, 階 kai) take sokuon but generally not rendaku. t- counters (頭 tō, 通 tsū) take sokuon only. This tool stores the irregular forms explicitly, but understanding the rule means you can guess correctly for counters not yet in the dictionary. To convert between kanji numerals like 三本 and Arabic 3本, kansuji-convert pairs naturally with this tool.

Industry and regional variation in counter usage

The same rabbit is counted 1 匹 in everyday speech but 1 羽 in zoology and rabbit-farming contexts — a quirk attributed to either the rabbit’s hopping motion or to Edo-period taboos around four-legged animal consumption (categorising them as birds dodged the taboo). Modern usage still splits along these industry lines: the Japanese Society for Laboratory Animal Sciences uses 匹 in its journals, while the Japan Rabbit Breeders Association uses 羽. Commercial usage varies too: たい焼き (fish-shaped pastry) appears as 1 匹, 1 個, or 1 枚 depending on the shop, with Tokyo favouring 個 and Osaka leaning toward 枚.

Contemporary Japanese has been simplifying counter usage by falling back on native-Japanese counting (1 つ hitotsu, 2 つ futatsu) for situations where the right counter is unclear. In business and polite speech this is sometimes considered safer than guessing wrong. Newspapers and official documents still enforce precise distinctions — イカ (squid) is 1 杯 ippai, 牛 (cow) is 1 頭 ittō, 豆腐 (tofu) is 1 丁 itchō — but daily speech increasingly uses the universal 1 つ. This tool surfaces the traditional counter; how strictly to apply it depends on register and audience. To extract candidate nouns from a longer draft, wakati-tokenize runs the same kuromoji POS tagger locally.