Skip to content

ALL-WRIT0003 · sentence_length_uniformity

Measure uniform sentence length without inferring authorship.

This is a deterministic rule for all languages. Read its implementation.

Extract the prose sections of a document, excluding code fences, comments, headings, tables, block quotations, and list items, and keep the sentences holding at least minimum_words words. For every section holding at least minimum_sentences of them, compute `100 * max(0, 1

  • MAD / mean)` over the sentence word counts, where MAD is the mean absolute deviation. Return the highest value any section reaches.

One hundred means every sentence in some section is the same length. Human prose varies its rhythm because the writer is thinking, so a section that does not vary is worth a second look. This is a measurement of rhythm and never a claim about who wrote the text, which is why the value comes back plain and a project policy decides what to do with it.

The finding names the section, its sentence count, and its exact bounded uniformity percentage. The value is the highest section uniformity in the document, where one hundred means equal sentence lengths and lower values mean more variation.

A section holding fewer than minimum_sentences qualifying sentences is skipped rather than measured, because a deviation over three sentences says nothing. Sentences shorter than minimum_words are dropped first, so a run of headings rendered as prose cannot drive the score. A document with no qualifying section at all measures zero rather than one hundred. Uniform prose is often deliberate in a reference manual, a translated text, or a templated report, so the value is evidence rather than a verdict.

Five sentences of the same word count return 100. Sentences of 5, 10, 15, 20, and 25 words have a mean of 15 and a mean absolute deviation of 6, so they return 60. A section holding four qualifying sentences is skipped under the default minimum_sentences, and a document of only such sections returns 0.