Guide

Parshandata searches the entire Tanakh with a query language that reads almost like plain English. You describe the pattern you are looking for and the engine finds or counts every match.

This page explains how queries as a whole are built. For the meaning of any individual keyword, open the keyboard on the search page and hover over its chip (or press and hold it on a touch screen) and a help card with its definition and examples of how it can be used will appear.

How it works

Parshandata runs on a custom query language that is built to feel like natural English. There is no fuzzy guessing and no AI in the loop — matching is exact and rule-based, so every count is precise. This makes it dependable for checking Masoretic notes or a fine point of grammar.

The text is broken into searchable layers: letters, vowels, cantillation marks (accents/trop), words, and verses. A query can search any one layer on its own. As you type, your search is checked for validity; when you run it, it is evaluated across every book you have selected, and the matches (or their count) are sent back.

The shape of a query

Almost every search follows one simple pattern:

count / find  every  unit  conditions…

• count returns a number; find returns the matching items themselves.
• The unit is what you are counting or finding — a word, a sentence, a letter, and so on.
• The optional conditions then narrow your search down.

count every word with esnachta→ 21,567 find every word with esnachta→ 21,567 Results

The first query above answers “how many”; the second lists the actual words. The counts shown throughout this guide are for the whole Tanakh with the default settings (Keri text, Ta'am Tachton) — change the selected books or modes and the numbers change with them.

What you can search for

The most common units are letter (a single consonant), word, and sentence (a pasuk / verse). There are also finer units — char (any character, including vowels and accents), cluster (a letter together with its attached marks), mark (any diacritic), and nekudah (a vowel) — and more.

count every word→ 305,452 count every sentence→ 23,203 count every letter→ 1,196,701

Adding conditions

Use with, without, that is, that has, beginning with, and ending with to filter:

count every word with 3 letters→ 72,352 find every word beginning with "א"→ 44,203 Results find every word ending with "ה"→ 34,013 Results

To match Hebrew text, put it in quotation marks. Pair it with ignoring mark to compare only the consonants, disregarding vowels and cantillation — so the bare skeleton "אלהים" matches every vocalized form:

count every word ignoring mark that is "אלהים"→ 681

Combining conditions: and / or

Join conditions with and and or:

count every word that has a "א" and has a "ב"→ 11,482 count every tropword with munach or mercha→ 73,042

Order of operations

A search is read left to right: count / find every <unit> first, then each condition is applied to that unit in turn. When conditions combine, a fixed order decides how they group — and knowing it lets you predict any search.

and requires both parts; or requires either. A word carrying both a mercha and a tipcha is rare —

count every word with mercha and tipcha→ 12

— while a word carrying either is common:

count every word with mercha or tipcha→ 68,862

When and and or are mixed, the and is resolved first. So this reads as kadmah, or (mercha and tipcha) — the and pair binds before the or:

count every word with kadmah or mercha and tipcha→ 6,938

Two more rules complete the picture. A relationship word — before, after, directly before — sits between two descriptions and joins the one on its left to the one on its right. And an article sets how many: a / an means “at least one such thing”, while the first / the last pick a specific one. Both get their own sections below.

When that default order is not what you mean, parentheses override it — which is the next section.

Parentheses change the meaning

This is the single most important thing to understand. When you mix and with or, the and is evaluated first — so these are not the same search:

Same words, same order — 40,162 versus 12, purely because of the parentheses. When a search mixes and and or, add parentheses to say exactly what you mean.

You don't have to add them all by hand — the TAB key in the search box rewrites your search with every grouping spelled out in parentheses (and back again), so you can see exactly how it will be read.

Indentation: the same grouping, more readable

Deeply nested parentheses get hard to read. Parshandata accepts indentation as an equivalent to parentheses — the TAB button in the search box converts a query between the two forms. This indented search means exactly the same as (munach or mercha) and tipcha above, and returns the same 12:

count every tropword with munach or mercha and tipcha→ 12

The indented lines form the group; the line that returns to the left margin (and tipcha) applies to the whole thing. Keep each level lined up — mismatched indentation is rejected.

Relationships between words

You can search for words in relation to one another with directly before, directly after, before, and after:

find every sentence with a disjunctive directly before an esnachta→ 10,206 Results

Each side can itself be a parenthesized condition — for example, every verse where a word bearing a kadmah sits immediately before a word bearing an azlah:

find every sentence with (a word with a kadmah) directly before (a word with an azlah)→ 3,350 Results

Narrowing the comparison: ignoring and using only

ignoring compares while disregarding a layer (we used ignoring mark above to ignore vowels and accents). using only does the reverse — it keeps only the named layer and strips the rest. This finds every distinct consonantal spelling in the text:

find every unique word using only letters→ 20,514 Results

Matching things to each other: the same / different

Most conditions ask whether a match has some named feature — an esnachta, a kamatz, a particular letter. the same asks something different: that several matches share a value, without you having to say which value. This finds every verse where two or more words are built on one root — polyptoton, a favourite device of biblical Hebrew:

count every sentence with at least 2 words with the same root→ 14,594

You name what has to match. It can be a property (the same root), a unit (begin with the same letter), a container (in the same word), or nothing at all, which compares the matches themselves (that are the same). Alliteration, three or more words from one initial:

count every sentence with at least 3 words that begin with the same letter→ 15,949

Doubled letters — gemination and scribal dittography — are a consecutive run that agrees:

count every word with at least 2 consecutive letters that are the same→ 8,879

And a word repeated inside one verse:

count every sentence with at least 2 words that are the same→ 4,354

different is the mirror image: instead of the largest group that agrees, it counts how many distinct values appear. Verses that draw on ten or more distinct initial letters:

count every sentence with at least 10 words that begin with different letters→ 4,381

Singular or plural changes what counts as matching. the same letter compares the letter sequence in order, so two words match only if they are spelled identically:

count every sentence with at least 2 words with the same letter→ 10,010

the same letters — plural — compares them in any order, which finds metathesis and anagrams. In Genesis 3:4, אֶל and לֹא use the same two letters reversed, so they match this way and not the other:

count every sentence with at least 2 words with the same letters→ 10,339

Properties have no order to lose, so their plural is just grammar — the same roots and the same root ask exactly the same question. The word the is optional throughout: same root reads the same as the same root.

Both words need a lower bound — at least 2, more than 3 — and not a plain number. 2 words with the same root is genuinely ambiguous: it could mean some root is shared by exactly two words, or that the largest group sharing a root is exactly two. Those are different questions with very different answers, so the search asks you to say which one you mean rather than guessing.

Grammar & morphology

Beyond letters and trop, you can search by grammar — the part of speech, verb stem and conjugation, and person, gender, number and state of each word. Open the keyboard and choose the Grammar tab for the full set of terms. Each term works on its own, as a kind of word:

count every noun→ 121,324

You can also phrase a term as something a word is. word that is a noun means the same as every noun:

count every word that is a noun→ 121,324

They also read as conditions, so you can combine grammar with everything else in the language. Here are the verbs in the hiphil stem:

count every verb with hiphil→ 9,401

…and they compose with and. Feminine plural words:

count every word with feminine and plural→ 7,071

The verb conjugations include wayyiqtol, the narrative past that drives biblical storytelling:

count every wayyiqtol→ 14,955

Because grammar terms are ordinary conditions, they combine with trop too. This counts the verbs that carry an esnachta, the mid-verse pause:

count every verb with esnachta→ 3,855

Grammar data is drawn from an external morphological analysis aligned to the text. A small fraction of words could not be aligned and carry no grammar; those simply never match a grammar term, rather than matching the wrong one. Grammar terms describe whole words — a sub-word unit (a letter or cluster) never matches one.

The grammar checker

As you type, Parshandata checks your search and shows whether it is well-formed. A green indicator means the grammar is valid; a red one means something is off, and the message points to roughly where the problem is — a line and column, with the nearby text quoted.

It checks form, not meaning: it confirms the search is built correctly, not that it will have matches. A perfectly valid search can simply describe something that never occurs, in which case it runs fine and returns zero.

Settings that shape every search

• Books — choose which books of Tanakh to search; every count is relative to this selection.
• Keri / Kesiv — search the text as it is read aloud (keri), as it is written (kesiv), or both.
• Ta'am Tachton / Ta'am Elyon — the Ten Commandments (Exodus 20 and Deuteronomy 5) carry two systems of cantillation. The lower accentuation (ta'am tachton) divides the passage into ordinary verses; the upper (ta'am elyon) divides it by commandment. This setting chooses which of the two — or both — your search reads.

Saving and sharing a search

Any search can be saved to a small file and re-uploaded later — that is exactly how the two examples on the About page work. Results can also be downloaded as a spreadsheet for further study.

A few searches to try

The last one is the eighty-letter search from the About page — the four verses of exactly eighty letters that open with "וכל".