Divine Discourses

Toward direct engagement with the Qur'an

Export: download the numbers

Fourteen precomputed tables, each as CSV and JSON: root frequencies and dispersion, root-pair association statistics, per-surah stats, verse lengths, formulas and formulaic density, centrality, rhyme summary, fawātiḥ, discursive pivots, computed structure and its tests, and theme-surah density. Every field is documented in the schema and data dictionary below. For the complete raw pipeline (morphology, per-root files, and every other dataset this site ships), see Datasets.

Generated deterministically by scripts/build-exports.mjs, which reads data/roots-summary.json, data/numbers.json, data/chronology.json, data/surah-profiles.json, data/surah-names.json, data/morphology/, and the association statistics from data/association/*.json (written by scripts/compute-association-stats.mjs). Run either script twice and the output is byte-identical. Both scripts and the full pipeline live in the public repository.

Leeds corpus v0.4, GPL. ~ Cairo 1924 for period fields.

ℹ Licensing and chronology of these files

Data derived from the Leeds corpus is GNU GPL licensed and redistributed under the same license; open either badge above for the full citation, and see NOTICE.md for the complete notice. Chronology-derived fields (period, revelation order) use the Cairo 1924 revelation order, whose periodization varies across scholarly chronologies. The export code itself is MIT licensed, like the rest of the site.

Downloads

Root frequencies ~

All 1,642 roots: raw count, overall normalized frequency (per 1,000 tokens), and per-period count and normalized frequency across the four revelation periods (Cairo 1924 / Nöldeke-Bell).

Root dispersion

All 1,642 roots: how evenly each is spread across the 114 surahs, weighted by surah size — Gries's DP, its Lijffijt & Gries normalization, Juilland's D, the count of surahs the root occurs in, and a dispersion-adjusted frequency. Four measures side by side rather than one, because the literature disagrees about which is best. Same figures as the Numbers table and the Roots root-detail panel.

Association pairs

Root-pair association statistics (shared verses, LLR, PMI, Dice): the union of every pair appearing in any root's top-25-by-LLR partner list. Same formulas and thresholds as the Roots statistical-associations panel; see the schema for the exact counting rule.

Per-surah stats ~

All 114 surahs: verse/token/root counts, vocabulary-diversity ratios, noun/verb percentage, and Cairo 1924 revelation order and period.

Verse lengths

Every verse in the corpus with its token length and revelation period.

Formulas

Every recurring 3-5 word sequence in the Qur'an, root-view and surface-view combined, with its first occurrence. Same counts as Formulas; full occurrence lists are in data/formulas-root.json and data/formulas-surface.json.

Centrality ~

All 1,642 roots: degree, weighted degree, betweenness, and eigenvector centrality over the root co-occurrence graph, each with its rank. Same figures as the root detail centrality panel.

Rhyme summary

All 114 surahs: verse-ending rhyme family count, dominant key and share, shift count, top refrain, and mean run length. Same figures as the Patterns rhyme panel.

Fawatih

The 29 surahs opening with isolated letters (al-muqatta'at) and which of the 14 distinct combinations each uses. No meaning is asserted for the letters.

Discursive pivots ~

137 verses mechanically flagged for opening with a temporal particle while sharing a root with the previous verse, a candidate marker of where a passage turns, not a scholar's reading.

Computed structure ~

Mechanically segmented sections for all 114 surahs (417 sections total), a changepoint model over lexical cohesion, never attributed to any named scholar. See the Dossier "Computed structure" panel.

Structure tests ~

Four block-level mirror-symmetry tests over the computed sections, one row per surah. 0 of 345 candidates reached significance after correction; see Patterns.

Theme-surah density ~

Sparse theme-by-surah matrix: for each surah, the themes whose root-family vocabulary clusters most densely in it (theme-root tokens per 1,000 surah tokens). See Themes.

Formulaic density ~

All 114 surahs: mean share of words covered by a recurring phrase (root and surface streams), tested against a length-matched resampling null, with a Benjamini-Hochberg survivor flag. Same figures as Formulas.

Schema and data dictionary

Every field, its unit, its counting rule, and its Verified/Nuanced status is documented in both a machine-readable schema and a prose data dictionary:

Everything at once, pinned to a version

One archive holds all fourteen tables in both formats, the schema, the data dictionary, and a citation file, so a download never travels without its provenance. The archive is named by release version and built deterministically: citing v1.2.0 names these exact bytes, and a new release adds a new archive rather than replacing this one. For toolchains that read standard packaging, the same tables are described as a Frictionless Data package and as Croissant JSON-LD.

Earlier releases stay downloadable, because a published citation points at the bytes it was written against: v1.0.0.

Preview: top 5 roots by frequency

Root Count Per 1,000 tokens
Loading…

Leeds corpus v0.4.

If you build something on this data, the contribution guide explains how to send it back, and Validation documents how claims earn their labels on this site.