Hebrew frequency analysis has two wrinkles: optional niqqud vowel points and the five letters that change shape at the end of a word. This tool counts how often each word appears, optionally folding those variants, and can cluster inflected forms by an approximate shoresh (three-letter root).
How it works
Tokens are split on whitespace and Hebrew and Latin punctuation. When normalisation is enabled the tool runs two passes before counting:
niqqud → remove vowel points and cantillation U+0591–U+05C7
finals → ך→כ ם→מ ן→נ ף→פ ץ→צFor shoresh grouping it strips frequent prefixes — the definite article ה and
the clitic letters ב כ ל מ ש ו — and common suffixes, then keeps up to three
core letters as an approximation of the triliteral root that underlies most
Hebrew vocabulary.
The five sofit (final) letters
Hebrew has five letters that take a distinct “final” form when they appear at the end of a word:
| Normal form | Final form | Letter name |
|---|---|---|
| כ | ך | Kaf |
| מ | ם | Mem |
| נ | ן | Nun |
| פ | ף | Pe |
| צ | ץ | Tsadi |
In Unicode, each pair occupies different code points (for example, Mem is U+05DE and Final Mem is U+05DD). Without final-letter normalisation, a word appearing mid-sentence would not match the same word appearing at sentence end, causing undercount. Enabling normalisation folds all five pairs before tallying.
What niqqud stripping does
Modern printed Hebrew and most digital text is unpointed. Liturgical texts, children’s books, and learner materials carry niqqud vowel points. When you run frequency analysis on pointed text without stripping nikud, a single word can appear as multiple tokens (e.g., one with full pointing, one partially pointed) that the counter sees as different strings. Stripping niqqud removes all combining marks in the U+0591–U+05C7 range before counting, merging these variants.
The shoresh (root) approximation
Most Hebrew content words derive from a triliteral root — a three-letter base
from which related words are built by inserting vowel patterns and adding
affixes. For example, the root כ-ת-ב (write) underlies:
כתב— wrote (verb)מכתב— letter (noun)כותב— writer / writing (participle)כתיבה— writing (gerund)נכתב— was written (passive)
Shoresh grouping strips the most common prefixes (ה, ו, ב, כ, ל, מ,
ש) and suffixes, then reduces the stem to three letters. This is an
approximation — a proper morphological analyser using a full lexicon would be
more accurate — but it meaningfully clusters related forms and reveals the
thematic distribution of a text.
Practical uses
- Vocabulary analysis: Find the highest-frequency words in a learner text to prioritise which vocabulary to teach or study.
- Concordance building: Identify all occurrences and frequency of key words in a biblical passage, sermon, or literary text.
- Authorship and style: Compare word distributions across texts or authors.
- Translation prep: Know which terms appear most often before starting a translation so you can decide on consistent renderings early.
Strip niqqud when working with pointed liturgical or learner text so it matches everyday unpointed spelling; keep niqqud if your study specifically concerns vowel patterns and pointing variants.