Roots
Every Arabic word grows from a root, a set of (usually three) consonants from which nouns, verbs, and adjectives are derived by fixed patterns. Pick a root to see its family of words, where it appears, and what appears alongside it; the corpus records 1,642 of them.
| # | Root | Arabic | Meaning | Period | Count |
|---|---|---|---|---|---|
| Loading… | |||||
Period bars (left to right): Early Meccan · Middle Meccan · Late Meccan · Medinan
● Leeds corpus v0.4. ~ English meanings are editorial orientation glosses, shown for the most frequent roots only.
Selected root detail
Occurrences ●
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4: total count of all token entries with this root across 114 morphology files (77,429 tokens in the corpus; 49,967 carry a root, the rest are particles, pronouns, and other unrooted forms).
Makki / Madani distribution ●
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4 morphology +
Quran.com Foundation API v4
/chapters revelation-place field. Surahs classified
makkah in the API are counted as Makki;
madinah as Madani.
Where it appears
Distribution by surah (114 surahs) ●
Each bar = one surah (1–114 left to right). Hover for surah number.
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4: summed token count for this root in each of the 114 surah morphology files.
Across the revelation timeline ●
Same counts as above, reordered by the Cairo 1924 revelation sequence and colored by period. Where does this root's usage concentrate as the revelation unfolds?
ℹ How this is computed
Per-surah counts (Leeds v0.4) placed at each surah's Cairo 1924 revelation position. The Cairo ordering is a convention (labeled Nuanced elsewhere on this site); and distribution over time does not by itself establish meaning or emphasis.
Frequency across revelation periods ~
ℹ How this is computed
Normalized frequency (occurrences per 1,000 tokens) in each of the four traditional revelation periods, by Cairo 1924 revelation order (Nöldeke-Bell four-period classification). Periodization varies across scholarly chronologies, so this trend is Nuanced, not a settled chronology.
Derived forms ●
One root yields many surface forms through Arabic morphological patterns. Grouped by Arabic form and part of speech (Leeds POS tag).
| Form | Part of speech | Count |
|---|
Group by lemma
Between the root and its inflected surface forms sits the lemma — the dictionary headword. Grouping this root's tokens by lemma and part of speech shows its actual derivational family: which nouns, adjectives, and verbs the root produces, and how often each is used. Every root-bearing token in the corpus carries a lemma, so nothing is omitted. ●
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4: grouped all token entries for this root by Arabic surface form and POS tag; sorted by frequency descending.
Roots that appear alongside ●
Roots that appear in the same verse most frequently (raw co-occurrence count).
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4: for each verse containing this root, recorded all other roots present in the same verse; ranked by co-occurrence frequency.
Verse references ●
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4: unique verse references (surah:ayah) where at least one token carries this root; sorted numerically.
Alongside, filtered
Co-occurring roots ~
ℹ How this is computed
Counted at verse level: two roots co-occur once for each
verse in which both are attested, regardless of order or
distance within the verse. Roots occurring more than 700
times across the whole corpus
(a-l-h, q-w-l, k-w-n, r-b-b, a-m-n, ʿ-l-m) are
excluded as counting partners: they are frequent enough to
co-occur with nearly everything, so including them would
make every root's list look the same. The 12 remaining
partners with the highest counts are listed below. This is a
separate, filtered view from the raw
Roots that appear alongside
list above. Co-occurrence
describes distribution in the text. It does not by itself
establish meaning.
The same filtered list as a map: closer to the center means nothing — line thickness and node size carry the count. Click a root to re-center on it.
Distinctive partners (PMI) ~
ℹ How this is computed
Pointwise mutual information (PMI) over the same verse-level attestation as the count list above: it asks how much more than chance two roots co-occur, not how often. A very frequent partner can top the count-sorted list while scoring low here — nothing distinguishes co-occurring with something that co-occurs with nearly everything. A rare, tightly paired root can do the opposite. This ranks a different question, not a more-correct version of the count list; both are kept side by side on purpose. Same function-word exclusion as the count list; a pair needs at least 3 shared verses before it's ranked at all — below that, a single shared verse between two rare roots produces an enormous but meaningless score.
The same distinctive partners as a map: node size and edge thickness carry the PMI score this time, not the raw count — a small node here can still be a high-count partner above. Click a root to re-center on it.
Co-occurring roots by period ~
ℹ How this is computed
The same verse-level co-occurrence count and function-word exclusion as "Co-occurring roots" above, computed separately within each of the four traditional revelation periods (Egyptian Standard order, Nöldeke-Bell four-period classification — the same periods used for the timeline above). Only periods where this root is attested at all are shown, top 6 partners each. A period with a smaller verse pool naturally yields smaller counts; this shows how a root's verse-mates shift or persist across the corpus's traditional chronology, not what any verse means.
Statistical associations ●
ℹ How this is computed
Computed from Leeds Quranic Arabic Corpus v0.4 morphology. Co-occurrence is counted at the verse level. Pairs require at least 5 shared verses. LLR is Dunning's log-likelihood ratio; PMI is pointwise mutual information; Dice is the Dice coefficient. Higher values indicate stronger distributional association.
In the network graph, node positions are precomputed offline (not laid out in the browser): the root sits at the center, its top partner roots sit on one of three rings by LLR rank (strongest partners closest), and each node's angle comes from a fixed hash of its own root letters, so a given partner always lands at the same angle. Node size reflects the root's raw corpus frequency; edge thickness reflects LLR. Click a node to open that root; hover or tab to it for the shared-verse count, LLR, PMI, and Dice. The graph plots the same partners the table below lists.
| Root | Shared verses | Strength of evidence (LLR) | Distinctiveness (PMI) | Overlap (Dice) |
|---|
These are statistical measures of word distribution in the text. They do not indicate thematic, exegetical, or theological relationships.
Network position ●
ℹ How this is computed
Computed by scripts/compute-centrality.mjs over
an undirected weighted graph built from
data/association/: nodes are the 1,642 roots,
edges are root pairs recorded there (each already meeting
the 5-shared-verse threshold), edge weight is LLR. Degree
and weighted degree are direct counts.
Betweenness uses Brandes' algorithm on the unweighted graph
(edge weight ignored for shortest paths). Eigenvector
centrality uses power iteration on the weighted graph
(tolerance 1e-9, max 1000 iterations, uniform starting
vector). Ranks are out of all 1,642 nodes, 1 = highest.
| How many partners (degree) | ||
| Combined partner strength (weighted degree) | ||
| Bridge position (betweenness) | ||
| Hub proximity (eigenvector) |
Centrality measures describe how a root is positioned within a network of statistical co-occurrence. They are strongly influenced by raw frequency: roots that appear often tend to score higher. These measures do not indicate thematic, exegetical, or theological importance.
Dispersion across the Qur'an ●
ℹ How this is computed
Computed by scripts/compute-dispersion.mjs over
the 114 surahs as corpus parts, weighted by each surah's
token count. Four measures reported side by side, since the
dispersion literature disagrees which is best: Gries's
Deviation of Proportions (DP) and its normalized form
(DPnorm), Juilland's D, and plain range. DPnorm 0 means the
root's occurrences track surah size exactly; DPnorm 1 means
total concentration in the smallest surah. Juilland's D runs
the other way (1 = perfectly even) and classically assumes
comparably-sized parts, which this corpus's 114 surahs (3 to
286 verses) are not — reported anyway for comparability with
the wider literature, not as the preferred measure.
Adjusted frequency discounts the raw count toward zero the
more clumped the root is.
| Surahs occurring in | ||
| Deviation of Proportions (DP) | ||
| Normalized DP (0 = evenly spread, 1 = maximally clumped) | ||
| Juilland's D (1 = perfectly even) | ||
| Dispersion-adjusted frequency | ||
Dispersion describes where a root's occurrences fall, not what they mean. It does not indicate thematic, exegetical, or theological importance.
Explore further
- Search this root's words on the Words page →
- Compare two roots' distributions side by side →
- Read a whole surah's profile on the Dossier page →
- For sourced lexicography, the canonical English reference is Lane's Arabic-English Lexicon (see Sources), and the morphology browser at corpus.quran.com.