The Hebrew alphabet

What is the Hebrew alphabet?

The Hebrew alphabet has 22 letters and is written from right to left. It is used for Hebrew, for Yiddish, Ladino and Judeo-Arabic, and it is the ancestor of the Arabic script and, by way of Greek, of the Latin one you are reading now. There are no capitals: a Hebrew letter has one form regardless of where it sits in a sentence.

It is an abjad rather than an alphabet in the strict sense, which means the letters are consonants and the vowels are mostly left for the reader to supply. Vowel points exist, a system of dots and dashes written above and below the letters, but outside prayer books, poetry and children's books they are usually omitted. That is why the same sequence of letters can be read several ways and why context does so much work.

Five letters do change shape, though not for grammar. Kaf, mem, nun, pe and tsadi each have a separate form used only at the end of a word. Those five final forms are the thing this page is built around, because they are where writing Hebrew goes wrong and because of what they do to the way the alphabet is encoded.

How to use this page

1. Read the table for the letter you need. Each row gives the ordinary form, the final form where there is one, the letter's name, a transliteration, its gematria value and both code points. The code points are what you need when a letter has to survive a paste into a system that stores bytes rather than shapes.
2. Paste a word into the checker. It rewrites the five letters into the right form for where they sit, in both directions: a final form stranded in the middle of a word is corrected back, and an ordinary form at the end is promoted. It also lists what it changed and where.
3. Read the gematria figure if that is what you came for. The tool totals the numeric values of the letters. A final form counts the same as its ordinary letter, which is the usual convention, and the section below explains the reading that differs.

Why the block holds 27 code points for 22 letters

Unicode gives Hebrew the range from U+05D0 to U+05EA, which is 27 code points. The alphabet has 22 letters. The extra five are the final forms, and they are not tucked away at the end of the block: each one sits immediately before the letter it belongs to. Final kaf is U+05DA and kaf is U+05DB. Final mem is U+05DD and mem is U+05DE. The pattern holds for all five.

So anyone generating this alphabet by walking the range gets 27 letters and is wrong five times over, with the finals silently mixed in among the ordinary letters as though they were separate members of the alphabet. The dataset behind this page is built letter by letter from the character database instead, and the tests assert the pairing rather than the count, because a count of 22 would also pass for a list that dropped five real letters and kept five finals.

This is the third alphabet on this site to have a structural trap of exactly this kind, and each one has a different shape. Greek has a hole: U+03A2 is unassigned because there is no capital final sigma, so 25 code points hold 24 capitals. Cyrillic has an outlier: the Russian alphabet is 33 letters but only 32 are contiguous, with Ё living at U+0401 among letters Russian does not use. Hebrew has interleaved extras. Three alphabets, three different ways for the obvious loop to be wrong.

The homoglyph problem that mostly is not here

The Greek and Cyrillic pages on this site are largely about lookalikes, because letters in those alphabets can pass for Latin ones and that is how spoofed domains and usernames work. Measured against the table Unicode publishes for exactly this purpose, Greek has 22 letters confusable with a Basic Latin letter and Cyrillic has 22 as well, which was a surprising dead heat when we measured it.

Hebrew has three. The script simply does not look like Latin: the letters sit on a different baseline, they are built from different strokes, and there are no capitals to imitate the uppercase forms that do most of the work in the other two scripts. Whatever else is difficult about Hebrew text, being mistaken for English is not it.

That makes the mistake worth guarding against a different one. In Greek the hazard is a character that is not what it appears to be; in Hebrew it is a character in the wrong form. A word ending in an ordinary kaf instead of a final kaf is not a security problem, it is simply misspelled, and it is the single most common error made by people typing Hebrew on a keyboard that does not do the substitution for them.

Honest limits

The checker works on letters only. It does not touch vowel points, cantillation marks or the geresh and gershayim, and it does not know whether the vowels you have written are the right ones. Those are separate systems layered on top of the letters, and correcting them needs a dictionary rather than a rule.

Its rule for where a word ends is that the next character is not a Hebrew letter. That is right for ordinary text, including text with punctuation or points, and it is wrong for the handful of cases where a word is interrupted by something that is technically a letter. It also cannot know that a final form is deliberate: some names, abbreviations and transliterations use one mid-word on purpose, and the tool will offer to correct those.

Gematria here counts a final form as equal to its ordinary letter, which is the standard convention. There is a well-known alternative in which the five finals continue the sequence from 500 to 900, and it is used for some purposes; the tool does not offer it, and any total you compare against a source using that system will differ. Neither reading is wrong, and this page states which one it used rather than leaving you to work it out.

Transliterations are the simplified style used on Israeli road signs and in passports. Academic transliteration distinguishes letters that this style merges, so a scholarly source will spell several of these differently and both will be correct for their purpose.

Why is it free?

Because it is a table and a small rule. The letter data ships with the page and the checker runs in your browser, so nothing you paste is uploaded, logged or stored anywhere.

So there is no account, no sign-up and nothing held back. The letters and their names come from the Unicode Character Database and the lookalikes from the table behind UTS 39, both regenerated by a script committed beside the page, along with 83 checks that verify the result.