All rules

One word spelled two ways

orthographic-varianton by default (info)Japanese / English

One word written two ways in one document, such as e-mail and email

Example

How to fix

Match the way the rest of the document writes it. If the team has decided, list it under prefer in chaff.yaml, and preferred-term checks that direction from then on.

Why it matters

Writing one word two ways in one document (e-mail and email, GitHub and Github) makes search treat them as different words and leaves the reader wondering whether two things are meant. chaff does not decide which way is right and keeps no list of words; it groups the document's own words, by reading in Japanese and by spelling without hyphens and case in English, and points at whichever way the document uses less. The case of the first letter is left out, since a sentence or a title gives it (Email, email), and so are words in capitals only, a name in capitals with an ending (SENDs) and the letters of an acronym picked out (CLImatology next to CLIPER). It says nothing when the minority is more than a third, since the document then uses both on purpose.

What it does not flag

The case of a word's first letter (Email, email; Long-Term, long-term) is not a spelling, nor is a word in capitals only (emphasis or an acronym), nor a name in capitals with an ending (SENDs, OPENed), nor the letters of an acronym picked out beside it (CLImatology next to CLIPER), nor a word inside a passage written in capitals (eGAIN in a disclaimer), nor a name coined from capitalised pieces that are not words (Medi-Cal beside medical). A name spelled with capitals (HBase) is still compared. In Japanese, words whose kanji and kana forms mean different things (成る and なる) are not compared, 方 and ほう are compared only before が or で (早いほうが), since elsewhere 方 is mostly the person (担当の方に), nor a label opening a sentence (the （答） of a Q&A), a bracketed spelling in a sentence that names other spellings (（くもり）、（曇）とも書く), kana たつ beside 立つ, since たつ is also time passing (経つ), a numeral (the 三 of 三十), kana written onto a pronoun (私みたような) or okurigana cut off a word (誂らえる); the head of a compound verb (出来上る), an adverb opening a sentence as a conjunction (――もっとも) and a katakana spelling that names a sense (モノ for goods) are compared apart from the word on its own (できる, 最も, もの), while the conjunction さらに、 and the adverb 更に are one word and are compared; in the legal genres each justice's separate opinion (裁判官甲の反対意見は次のとおりである。) is compared only within itself; the judging みる (〜とみられる) and いえる (〜とはいえない) are compared apart from looking and saying (資料を見る, 一概には言えない), and in the legal genres neither are 時 and とき, 物 and もの or 在る and ある, which legal drafting writes apart. A noun and a verb (払込み and 払い込む) are different words, even where the analyser reads the noun as the verb. Hyphenated words that are another word (re-sign, resign), part of an address (github.com), anything in quotation marks (single ones too, as an opinion quotes a statute over several sentences), the title of an outside page quoted by a link at a line's head, a Japanese reading in brackets right after a word (HTTP（えいちてぃーてぃーぴー）), a Japanese company name written with its form (青葉硝子㈱) and the team's names are not compared. Nothing is reported when the minority is more than a third. The literature preset turns it off, since old spelling belongs to its time. Nor is a kanji spelling that reads as two words (行っ, read as 行く or 行う; 様 before を, read さま) compared with the kana of the reading the tokenizer picked.

Turning it off or relaxing it

This page is built from the rule's definition file:packages/chaff/rules/orthographic-variant.yaml