Paragraphs that lean toward one text

Urantia paragraphs whose closest text stands out above their mean across texts, ranked by that gap. A lean counts only when the text has the paragraph in its top 10% and both embedding models pick the same text. Filter with corpus (a corpus name or bible).

Scores come from the nine world religions texts and the World English Bible. For each text, a paragraph's best match becomes a percentile within that text, after an adjustment for paragraph length. The texts are not independent votes: a paragraph that sounds devotional tends to score well in all of them. These are measures of closeness in meaning by an embedding model, not claims about the teachings.