ALL-WRIT0005 · sentence_opener_concentration

Measure the share of sentences using one dominant opening word.

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

Definition

Take the first word of every sentence in each prose section, fold its case, and drop the words listed in ignored_openers. In each section holding at least minimum_sentences of the remaining openers, divide the count of the most frequent opener by the number of openers and state it as a percentage. Return the highest value any section reaches.

An opener repeated across most sentences is a rhythm a reader hears, and it is usually a sign that each sentence was started rather than continued. Like the length measures this reports a share and claims nothing about who wrote the text.

Evidence

The finding names the dominant opener, how many sentences it opens, how many openers the section holds, and the resulting percentage. Ties are broken alphabetically so two runs over the same text report the same word. The value is the highest section concentration in the document, as a percentage of its eligible sentences.

Exceptions

A section holding fewer than minimum_sentences eligible openers is skipped rather than measured. ignored_openers drops the words whose repetition means nothing, which is a and i by default, and a project working in another language states its own. Repetition can be deliberate anaphora or required terminology, and first-person prose repeats its subject by nature, so the value is evidence a person reads rather than a verdict.

Examples

Where This opens three of six eligible sentences, the section returns 50. Six sentences opening with six distinct words return about 16.67. A section holding five eligible openers is skipped under the default minimum_sentences, and a document of only such sections returns 0.

References

• Cites “Vale AI Tells”, experimental SentenceStartRepetition rule. Open reference
• Cites “Do LLMs Write Like Humans”. Open reference