Skip to main content
All docs

The prose engine

Writing in your language

Which languages Meridian and the Sentinel read, what they measure in each one, and which checks are not available.

Meridian and the Sentinel read ten languages: English, German, Spanish, French, Italian, Portuguese, Russian, Japanese, Korean and Chinese. Set the language on a piece and both engines switch to that language’s own rules rather than translating English ones.

If a reading looks wrong for how your language is written, tell us through Support.

Readability formulas

Every language uses a readability formula built for it:

  • German: Lesbarkeitsindex (Amstad)
  • Spanish: Lecturabilidad (Fernández Huerta)
  • French: Indice de lisibilité (Kandel-Moles)
  • Italian: Indice Gulpease, which counts letters rather than syllables
  • Portuguese: Índice de Legibilidade (Flesch adaptado)
  • Russian: Индекс удобочитаемости (Оборнева)
  • Korean: 국어 이독성 공식, which reports a school grade rather than an ease score

The reading you get names the formula it used.

Your dialogue is read as dialogue

German uses low and high quotes, French and Russian use guillemets, and Spanish, Portuguese and Italian often open the line with a dash instead. Both engines read all of those as speech.

Every check that excludes dialogue uses this: point of view, tense, sentence fragments, and the Sentinel’s continuity scans all look at your narration and skip what your characters say.

Tense follows your grammar

Each language is read with its own grammar rules.

When the engines cannot tell, they report nothing.

Checks that are not available in every language

Japanese and Chinese are written without spaces between words, so checks that count words are not reported for them: sentence length, vocabulary variety, paragraph rhythm. The report shows which checks ran.

Chinese marks aspect instead of tense, so tense consistency is not reported for Chinese.

When the Sentinel cannot read a language

A continuity check that cannot read your language does not run, and the report lists which checks did not run and why.

Dialogue attribution advice

Attribution variety is measured in every language. Advice about it appears only in the languages whose writing tradition holds that rule. A language that has not been reviewed gets the measurements and no advice.

Tell us what reads wrong

If a reading says something about your prose that a writer in your language would not say, send it through support, in whatever language you write in.

Still stuck? Head back to Support to report a bug or reach the team.