Provenance
The sources
- Arabic text
- quran-uthmani, distributed by Tanzil.net.
- Recitation (qiraʾa)
- Ḥafṣ ʿan ʿĀṣim — the reading of the great majority of printed muṣḥafs today.
- Morphology
- Quranic Arabic Corpus — root, lemma, and part-of-speech tagging for every word.
- Verse numbering
- Canonical (Kūfan) numbering, the scheme used by the standard muṣḥaf: 6,236 verses in 114 sūrahs.
- Translations
- 31 translations in 29 languages, all from Tanzil.net, each credited to its translator.
- Roots
- 1,651 distinct roots derived from the morphology.
The rule
Three columns, never blended
Every claim on Tartīl belongs to exactly one of three columns, and the columns are never quietly mixed. This is the discipline the whole project rests on.
Computed
Directly counted or parsed from the corpus above — “this root occurs 140 times in 128 verses.” Reproducible: same text, same morphology, same number. If you disagree with a computed figure, you can check it.
Statistical
The output of a model — co-occurrence strength, centrality, clustering — always published with its null model and its corrections. A statistic is a measurement under assumptions, never a verdict about meaning.
Interpretation
Meaning — glosses, themes, tafsīr — always carrying its source and how strongly that source holds. Where the tradition genuinely forks, both readings are shown; we do not flatten one into a “fact.”
One consequence worth stating plainly: a semantic label — “this verse is about mercy” — is interpretation, even when a machine produced it. Automatic tagging is quiet tafsīr, and we mark it as such rather than dressing it up as computation.
Counting
What a number actually counts
Most disagreements about Qurʾānic word counts are not disagreements about the text — they are unstated differences in the counting unit. Tartīl uses these definitions consistently:
- Occurrences — the number of word tokens assigned to a root. A root appearing twice in one verse counts twice.
- Verses — the number of distinct āyāt containing at least one such token. Always less than or equal to occurrences.
- Forms — the distinct lemmas derived from a root, each with its own count.
Where a page shows a figure, it says which of these it is. A bare number with no unit is not a fact we are willing to publish.
Honest limits
Where our numbers will differ
Another site may publish a different count for the same root, and both can be correct under their own declarations. The differences come from:
- The qiraʾa. Different canonical readings spell some words differently. Ours is Ḥafṣ ʿan ʿĀṣim; a count from a Warsh text is not comparable to ours without saying so.
- The morphological analysis. Which root a word derives from is a scholarly judgement. Reasonable analysts assign some words differently, and the counts move with them.
- Proper nouns and particles. Whether a name or a function word is credited to a triliteral root at all is a convention, not a discovery.
This is also why Tartīl is not a numerology engine. We publish counts to enable honest study — and, where a striking “miracle count” turns out to depend on an undeclared edition or a hand-picked counting rule, to deflate it rather than repeat it.
Interpretation
How sources are weighted
For anything in the interpretation column, Tartīl records not only the claim but the kind of warrant behind it. The strongest is the Qurʾān explaining the Qurʾān — one verse clarifying another — which outranks any single transmitted report or opinion. Transmitted narrations carry their grading with them, and a weak report may be quoted but never made load-bearing.
Where a widely held reading and the statistics point in different directions, we surface the gap deliberately instead of hiding it. A root can be statistically central and doctrinally light, or rare and weighty. That divergence is itself a finding, not an error to be smoothed away.
Read further: