What the checker flags: a legend of issue types

Japanese reference books open with a few pages called 凡例 (hanrei), the key that tells you how to read the entries. This page does the same job for the checker on the front page. Every finding in the result list carries an issue type, and each type is explained below: what it looks like, why it matters, how to decide, and an example with a fix.

One point applies to every finding. The checker works from rules and word lists: it looks for patterns in the characters you paste and does not understand what your sentence means. It will flag things that are fine, and it will miss many real problems, such as a wrong homophone that is also a real word or a number that does not match the source. Read each finding as a question, not a verdict. The guide to using the checker covers the limits in more detail.

The issue types fall into three levels, matching the three rule groups you can switch off:

Basic: likely errors

Spelling and conversion slips

The checker has curated entries for common mistakes such as シュミレーション → シミュレーション, コミニケーション → コミュニケーション, うる覚え → うろ覚え and 人事異同 → 人事異動. These are finite lists, not a general spellchecker dictionary. A proper name or deliberate quotation can differ from the usual spelling.

A specific replacement has a button, but it is never applied automatically. Confirm the intended word before using it. The checker no longer treats disputed expressions such as 的を得る or 汚名挽回 as definite conversion errors, and it does not change キャノン to a company's spelling merely because the characters match.

Adjective endings

For selected common い-adjectives, the checker looks for missing small っ and malformed past endings: おいしかた → おいしかった, 楽しいでした → 楽しかったです and 高くないでした → 高くなかったです. It also knows that いい has the past form よかった, while な-adjectives such as きれい take だった or でした.

For 高いだ, the choice between 高い and 高いです depends on the intended voice, so the tool asks you to review rather than choosing for you. 高いだろう is a valid construction and is not treated as this error. The supported vocabulary is limited; a missing flag is not proof that an unfamiliar adjective is correctly formed.

Verb and polite endings

The checker uses an explicit set of common verbs to review forms such as 読むます → 読みます, 行きませんだった → 行きませんでした and 食べますた → 食べました. It also checks selected plain past slips and constructions such as 行きますつもり, where a plain form is normally used before つもり.

A fragment such as いきまて does not uniquely identify the intended ending. It might have been いきます, いきました or いきまして. The checker lists alternatives without inventing a tense. Valid plain forms inside a polite sentence, including 行くつもりです and 雨が降るかもしれません, are not errors simply because they lack ます.

A particle next to a quantity

Sequences such as はっ二個 or をっ三冊 can indicate an accidental small っ between a particle and a quantity. The tool highlights the join and asks you to inspect the whole phrase. It cannot infer the correct word order from this sequence alone. For example, “two apples” can be りんごを二個, but は can also be a valid topic marker in another sentence.

Unbalanced brackets and quotation marks

Opening and closing marks are checked as pairs, including （）, 「」, 『』, 【】 and ordinary parentheses. A missing or mismatched partner gets a review note. An excerpt may legitimately omit the opening or closing mark, so the tool does not insert one automatically.

Doubled particles

Selected particles repeated in a row: をを, がが, でで and にに. Often an editing leftover: you rewrote half a sentence and the old particle stayed next to the new one.

How to decide: read the characters around the flag first, because some sequences are legitimate. The checker does not flag のの or はは; it also excludes common なにに and かにに sequences. A finite rule cannot identify every word boundary, so other false alarms remain possible.

Example: 資料をを送ります → 資料を送ります ("I will send the documents").

Doubled punctuation

Repeated full stops (。。), repeated commas (、、), or a comma immediately followed by a full stop (、。). In body text these are slips from pasting or fast typing. Repeated ！ or ？ may be intentional in casual text or dialogue and are not reported by this rule.

Example: 確認しました。。→ 確認しました。

Readability: things that slow readers down

Very long sentences and too many commas

Japanese can chain clauses with 〜て, 〜が, 〜し and 〜ので almost indefinitely, and the main verb comes last, so the reader carries every clause until the predicate arrives. The checker flags sentences above a length threshold and sentences with many 読点 (tōten, the comma 、).

How to decide: as a rough guide, a sentence past about 100 characters, or with more than four or five commas, deserves a second look. Legal and technical text is often long by convention; explanations and instructions gain most from cutting. Split at a connective: 〜が becomes 。ただ、 and 〜ため becomes 。そのため、.

Before: 新しいシステムは四月から導入される予定ですが、データの移行作業がまだ終わっておらず、担当者も不足しているため、開始日が遅れる可能性があり、利用者への案内も準備中です。

After: 新しいシステムは四月から導入される予定です。ただ、データの移行作業がまだ終わっておらず、担当者も不足しています。そのため、開始日が遅れる可能性があります。利用者への案内は準備中です。

Double negatives

Constructions such as 〜ないわけではない ("it is not that ... not"), 〜なくはない and 〜ないとは言えない. Japanese uses them to soften: 行けなくはない ("it's not as if I can't go") is a hedged yes.

How to decide: in conversation and essays the softening is often the point. In instructions, rules and anything a reader must act on, a double negative makes them stop and work it out, so state it positively.

Example: 条件を満たさない場合でも申請できないわけではありません → 条件を満たさない場合でも、申請できることがあります ("even if you do not meet the conditions, you may be able to apply").

Machine-dependent characters

The 機種依存文字 (kishu izon moji): circled numbers such as ①, enclosed abbreviations such as ㈱ (株式会社, "Co., Ltd."), squeezed units such as ㌔, and single-character Roman numerals such as Ⅳ. They sat outside the old JIS X 0208 standard and were added by individual computer makers, so they could show as a different character on another machine or turn into garbage in email.

Most devices now display them through Unicode, but some older systems and forms still reject them, screen readers may read them oddly, a search for 株式会社 will not find ㈱, and ㌔ could be kilograms or kilometres.

How to decide: print with a known font may be fine. For email, forms, the web and anything read by software, replace them.

Fixes: ① → (1); ㈱ → 株式会社 or （株）; ㌔ → キロ, kg or km; Ⅳ → IV in ordinary Latin letters.

Style: consistency and register

Historical spelling and formal-writing preferences

きのふ and けふ are historical spellings of きのう and きょう. They are valid in historical or literary writing. In everyday mode, the checker offers the modern form as a style choice; literary mode suppresses this suggestion. It does not attempt to modernize every historical spelling.

Formal mode adds a small set of conventional-expression preferences. Ra-nuki forms such as 見れる and redundancies such as まず最初に are also style suggestions, not blanket declarations that speech or dialect is wrong. Literary mode suppresses these preferences. Check house style and context before changing them.

Mixed polite and plain sentence endings

Japanese prose uses either the polite style, 敬体 (keitai, ending in です and ます), or the plain style, 常体 (jōtai, ending in だ and である). Switching between them in one text reads as a break in voice.

How to decide: the checker looks only at sentence endings, so it can be fooled. Quoted speech keeps its own ending, headings and bullet points often end on a noun (体言止め, taigen-dome) and belong to neither style, and a plain form inside a polite sentence, as in 便利だと思います, is normal. What matters is that the main narrative stays in one style.

Example: 本製品は軽量です。持ち運びにも便利である。 → 本製品は軽量です。持ち運びにも便利です。 ("This product is light. It is also easy to carry.")

Half-width katakana

Narrow single-byte katakana: ｶﾀｶﾅ instead of カタカナ. They come from early character sets (JIS X 0201) and still arrive in text copied from older systems. Their voicing marks are separate characters, so ﾃﾞ is two characters where デ is one, which breaks searching and sorting, and they look cramped beside full-width kanji.

How to decide: convert them, unless a form or system specifically asks for half-width kana.

Example: ﾃﾞｰﾀを入力 → データを入力 ("enter the data").

Mixed full-width and half-width letters and digits

One text using both Ｗｅｂ and Web, or ２０ and 20. Neither width is inherently wrong; mixing widths may be intentional for different kinds of content. A common house rule for horizontal text is half-width Latin letters and digits, while some publications keep full-width digits in vertical text. Spaces count too: the full-width space (　) and the ordinary space look alike and behave differently.

How to decide: pick one rule for the document and apply it with search and replace.

Example: 申込は２０件、キャンセルは5件 → 申込は20件、キャンセルは5件.

Inconsistent spellings

One word written two ways in the same text, called 表記ゆれ (hyōki yure). The checker keeps a list of known pairs and flags a pair only when both forms appear: サーバ / サーバー and コンピュータ / コンピューター (the final long vowel mark), 下さい / ください, 出来る / できる.

How to decide: either form can be right; what matters is choosing one. For loanwords, the cabinet notice 外来語の表記 (1991) writes long vowels with ー as a rule, while technical writing long dropped the final ー on longer words, and many organisations now use the long form. For ください, the usual convention, also followed in official writing, is kana for the auxiliary (確認してください, "please check") and kanji for the verb meaning "give" (資料を下さい, "please give me the documents"). Record your choices on a short style sheet; our post on hyōki yure covers longer documents.

Example: サーバーを再起動し、サーバの状態を確認する → サーバーを再起動し、サーバーの状態を確認する ("restart the server and check its status").

What this legend does not cover

This is a finite set of checks, not exhaustive Japanese proofreading. URLs, email addresses and backtick code are skipped. Wording inside paired Japanese or curly quotation marks is skipped unless you enable that option. Homophones that are both real words (以外 and 意外, 保証 and 保障), the right level of keigo, figures against the source and plain truth all need a person. When a finding seems wrong and you know why, trust your reasoning over the flag.

Grammar references

The Tokyo University of Foreign Studies adjective tables show the distinction between い- and な-adjective forms. Its へ lesson explains why both スーパーへ行く and スーパーに行く are possible. The Japan Foundation explanation of つもり illustrates plain forms before つもり. These references support the explanations; they are not a certification of this checker.