2. Script and spelling

This chapter covers how Modern Hebrew is written. It describes the 22 letters and their five final forms, the vowel points (niqqud) and the other diacritics, and the two spelling systems: pointed spelling, and the full spelling without points (ktiv male) used in almost every everyday text. It covers the Academy of the Hebrew Language's rules for full spelling, the geresh and gershayim, abbreviations, numbers written with letters, punctuation, Unicode and right-to-left layout, and how ambiguous unpointed text is in practice. It closes with the romanization standards and the transliteration used throughout this manual.

Index: Modern Hebrew (he): a reference manual for Mova · previous: 1. Overview: Modern Hebrew · next: 3. Phonology · related: Nouns, adjectives, pronouns and numerals, 11. Modern Hebrew in Universal Dependencies

2.1 The alphabet

Modern Hebrew uses the Hebrew "square script" (ktav ashuri, כְּתָב אַשּׁוּרִי), which descends from the Imperial Aramaic script. It is an abjad: the letters write consonants, and vowels are either left out or shown by a few letters doing double duty (matres lectionis, אִמּוֹת קְרִיאָה, imot kri'a), or by the optional points. It is written right to left. There is no upper or lower case. Words are separated by spaces, but the article, some prepositions and conjunctions, and the relative particle are written joined to the next word (see §2.10 and 11. Modern Hebrew in Universal Dependencies).

The table gives each letter's name as it is said today (the Academy's spelling of the name in brackets), its value in standard Israeli speech, its traditional numerical value, and its final form where it has one. "Dagesh" means the letter written with the inner dot (§2.3.1).

Sources for the modern values: Laufer (1999), Bolozky (1997), Asherov & Cohen (2019); the overview in Wikipedia "Hebrew alphabet" and "Modern Hebrew phonology". The historical column is a simplification of the Tiberian reading tradition as described by Khan (2020). 3. Phonology gives the sound system in full.

2.1.1 Final forms

Five letters have a separate shape at the end of a word: כ→ך, מ→ם, נ→ן, פ→ף, צ→ץ (mnemonic mantsepakh, מנצפ״ך). The final form is a matter of position, not of sound. ך is [χ] (khaf sofit) and is written with a dagesh, ךּ [k], only in a few words and in pointed text. ף is [f] in native words. Loanwords and slang ending in [p] or [ts]/[tʃ] are a known problem area:

• [p] at the end. Some writers use the ordinary final ף, and the reader has to know it is [p]. Others keep a non-final פ at the end of the word (שופ or שוף "shop", צ׳יפ or צ׳יף "chip"), and this is common in brand names. Pointed text can write ףּ (final pe with dagesh). Expect both spellings in data.
• [tʃ] at the end. Final [tʃ] is written either as ץ׳ or as a non-final צ׳ (e.g. סנדוויץ׳ / סנדוויצ׳ "sandwich"). Usage varies.

The variation is real and should be expected in data. A tokenizer or spell checker must not "correct" a non-final פ or צ at the end of a word in loanwords.

2.1.2 Letters with two values and letters with one value

Because historical sounds have merged (see 03-phonology.md §3.2.2), several letters share a sound, and three letters have two sounds:

In unpointed text, these letters cannot be told apart, so the reader must know the word. This is one main source of ambiguity (§2.9).

2.2 The vowel points (niqqud)

The points are a system of dots and dashes set below, inside or above the letters. The Tiberian Masoretes created them in the 8th–10th centuries CE to record a reading tradition of the Bible that had more vowel distinctions than Modern Hebrew (Khan 2020). Modern speech has five vowels (/a e i o u/, 3. Phonology), so several signs are now pronounced the same.

Today pointed text is used in:

• poetry and song lyrics;
• books for small children and early readers, and materials for learners (ulpan);
• the Bible and the prayer book;
• dictionaries, and occasionally to disambiguate a single word in otherwise unpointed text (niqqud ezer, "helping points");
• some official notices and signs aimed at new immigrants.

Everyday prose (newspapers, books for adults, the web, messages) is unpointed.

2.2.1 Vowel signs and their modern values

The table shows each sign on the base letter ב (shown in the "Shape" column), its traditional name, its traditional class, and its value in standard Israeli speech.

Under modern pronunciation, then, the vowel signs collapse into five groups: patakh + kamats gadol + khataf patakh = [a]; tsere + segol + khataf segol (+ some shva) = [e]; khirik = [i]; kholam + kamats katan + khataf kamats = [o]; kubuts + shuruk = [u]. The difference between kamats and patakh, or tsere and segol, survives only in spelling, and in the spelling rules that depend on it (e.g. tsere is a reason to write yod in some words; §2.4.4). Sources: Laufer (1999); Wikibooks Dikduk ha-Ivrit ha-Hadasha, "Vowels" page (Hebrew Wikibooks); Coffin & Bolozky (2005, ch. 1).

2.2.2 Shva: na and nakh

A shva under a letter means that the letter has no full vowel. The grammars distinguish:

• shva nakh (שְׁוָא נָח, "resting"): silent, ending a syllable, e.g. the shva under ס in יִסְגֹּר yisgor "he will close";
• shva na (שְׁוָא נָע, "moving"): in the Tiberian tradition a short vowel, e.g. under the first letter of a word, under a letter after another shva, under a doubled (dagesh forte) letter.

Modern speech does not follow the traditional rules consistently. A shva is said as [e] where the speaker needs a vowel to make the word pronounceable or where morphology requires it, and is silent elsewhere. Some typical outcomes (Bolozky 1997; Laufer 1999):

Rules of thumb for Modern Hebrew:

1. A shva in a prefix (בְּ, כְּ, לְ, וְ, the participle prefix מְ) is said [e]: be-, ke-, le-, ve-, me-.
2. A shva on the first of two consonants that cannot form an onset is said [e] (§3.6): lerakhem, not *lrakhem. The forbidden initial clusters include a first sonorant such as ר, ל, נ, מ or י in many words, and two consonants that are identical or differ only in voicing.
3. Elsewhere at the start of a word, a shva is usually silent: zman, shmone, kfar, sfarim, tmuna.
4. Inside a word, a shva after a closed syllable is silent: mikhtav, yikhtov.

The UD treebanks of Modern Hebrew are unpointed, so none of this is visible in their FORM fields. A speech or text-to-speech system, or a model that uses the points, has to deal with it.

2.3 Other diacritics

2.3.1 Dagesh

The dagesh (דָּגֵשׁ) is a dot inside a letter. In the Tiberian system it marked two things:

• dagesh kal ("light"): the stop value of ב ג ד כ פ ת (the begadkefat letters) at the start of a word or after a closed syllable;
• dagesh khazak ("strong"): a doubled (geminate) consonant, e.g. in the pi'el pattern (דִּבֵּר diber "spoke") or from an assimilated נ (יִפֹּל yipol "will fall").

In modern pronunciation, gemination is gone, and the stop/spirant split is audible only for בּ/ב, כּ/כ, פּ/פ. For the other letters, the dagesh is a spelling sign. In ordinary unpointed text, even the letters with two values carry no dagesh, so כ can be [k] or [χ] and פ can be [p] or [f] (but see §2.4.3: final ף is always [f] in native words).

Two other dot signs take the same place:

• shuruk וּ = [u] (a vav with a dot, which is a vowel, not dagesh);
• mappiq (מַפִּיק) in a final ה: הּ marks the ה as a consonant, not a vowel letter. Today it appears mainly in the 3rd person feminine possessive suffix: סִפְרָהּ sifra(h) "her book" vs. סִפְרָה sifra "digit, numeral"; לָהּ la(h) "to her" vs. לָה (no such word). The [h] is said in careful or formal speech and usually not in casual speech (Asherov & Cohen 2019), so the two forms are often homophones.

2.3.2 Shin dot and sin dot

שׁ (dot top right) is shin [ʃ]; שׂ (dot top left) is sin [s]. In unpointed text both are ש. Most ש in Modern Hebrew are shin. Common words with sin include שָׂם sam "put", שָׂמֵחַ saméakh "happy", יִשְׂרָאֵל Yisra'el "Israel", שָׂפָה safa "language", עֶשֶׂר éser "ten", שָׂדֶה sade "field" and מַעֲשֶׂה ma'ase "deed".

2.3.3 Rafe, meteg and the accents

The rafe (רָפֶה), a bar above a letter, marks the spirant value, and appears in some older printing and in Yiddish. The meteg (מֶתֶג), a small vertical bar under a letter, marks secondary stress and some vowel lengths. The cantillation accents (te'amim, טְעָמִים), which give stress and musical phrasing in the Bible, play no role in Modern Hebrew writing. They are encoded in Unicode (U+0591–U+05AF) and appear in Bible text in the same corpora that hold Modern Hebrew, so a pipeline should strip them. The same applies to the special characters U+05C0 (paseq), U+05C3 (sof pasuq) and U+05C6 (nun hafukha).

2.4 Spelling without points: ktiv male

2.4.1 Two systems

Hebrew has two spelling traditions:

• ktiv menukad (כְּתִיב מְנֻקָּד, "pointed spelling"), also called ktiv khaser (כְּתִיב חָסֵר, "defective spelling") when it is unpointed. Words are spelled with the minimum of vowel letters, as in the Bible and pointed texts, and the points carry the vowels. Example: שֻׁלְחָן "table", דִּבֵּר "spoke", חָכְמָה "wisdom".
• ktiv male (כְּתִיב מָלֵא, "full spelling"), the official name being ha-ktiv she-eino menukad, "spelling without points". Extra ו and י are added to help the reader where there are no points: שולחן, דיבר, חוכמה. Consonantal ו and י are doubled to tell them apart from vowel letters: תקווה tikva "hope", עניין inyan "matter".

All modern unpointed text is in some version of ktiv male. In practice, it is a mixed system: the Academy's rules are recommendations, newspapers and publishers have house styles, and individual writers vary. Both לחשוב and לחשב lakhshov "to think", מילה and מלה mila "word", and אשה and אישה isha "woman" are frequent. A lexicon or tokenizer for Modern Hebrew therefore needs spelling variants linked to one lemma. Hspell, for example, follows the Academy's rules strictly and accepts only the spellings they allow (Har'El & Kenigsberg, Hspell documentation), while the UD treebanks keep the spelling of the source text.

2.4.2 History of the rules

• 1947 (Tishrei 5708): the Hebrew Language Committee (Va'ad ha-Lashon), the Academy's forerunner, publishes "rules for spelling without points".
• 1968–1993: the Academy of the Hebrew Language, founded in 1953, revises and republishes the rules, and they become standard in education.
• 2002: the Academy publishes the rules together with the punctuation rules (Klalei ha-pisuk ve-klalei ha-ktiv she-eino menukad).
• 2017 (Sivan 5777): the Academy publishes revised rules of full spelling (Klalei ha-ktiv ha-male). The best-known change is the end of the 1947 exception that wrote אשה isha "woman" without yod: the standard spelling is now אישה. Other details changed as well.

Sources: Academy of the Hebrew Language, "Klalei ha-ktiv she-eino menukad / ha-ktiv ha-male" (hebrew-academy.org.il, the decisions section, and the PDF of the 2017 rules); Hebrew Wikipedia, "כתיב חסר ניקוד" (accessed 2026-10-17). The Academy's site refused automatic fetching on 2026-10-17 (HTTP 403), so the summary below follows the published rules as they are summarized in the Hebrew Wikipedia article and the grammars. Check the Academy PDF before relying on a fine detail.

2.4.3 The rules in outline

The rules are given by the vowel in pointed spelling, so a writer has to know the pointed form. That is why native writers also vary.

(1) /u/ (kubuts or shuruk) is always written with ו.

(2) /o/ (kholam, and kamats katan) is written with ו, with a few exceptions.

The main exceptions are כָּל kol "all", written כל, and words where /o/ is already shown by א, such as רֹאשׁ → ראש rosh "head", לֹא → לא lo "no, not", זֹאת → זאת zot "this (F)" and צֹאן → צאן tson "sheep".

(3) /i/ (khirik) is written with י, except mainly where a shva follows (a closed syllable) or in certain fixed small words and prefixes.

(4) /e/ (tsere, segol) is normally written without י. The common exceptions are a group of words that traditionally have a yod (תֵּימָן teman "Yemen" → תימן, מֵיתָר meitar "string" → מיתר), and forms where the tsere comes from compensatory lengthening before a guttural or ר (pi'el forms whose middle root letter cannot double): תֵּאֵר te'er "described" → תיאר, תֵּאוּר te'ur "description" → תיאור, פֵּרוּשׁ perush "commentary" → פירוש, בֵּרֵךְ berekh "blessed" → בירך. Here the full spelling treats the tsere as if it were the pi'el khirik.

(5) Consonantal ו [v] is doubled (וו) inside a word, but not at the start or end of the word, and not when the result would have three ו in a row.

After a prefix, a word-initial ו is doubled (הַוֶּרֶד → הוורד ha-véred "the rose"), except after the conjunction ו: וְוֶרֶד → וורד ve-véred "and a rose", not *ווורד.

(6) Consonantal י [j] is doubled (יי) inside a word and often at the end, but not at the start of a word, not after a prefix, and not next to a vowel letter ו or י.

The rule for the end of a word has some exceptions, in short frequent words: חַי khay "alive" → חי, דַּי day "enough" → די, שַׁי shay "gift" → שי, and מַיִם máyim "water" → מים. But שָׁמַיִם → שמיים "sky".

(7) No more than two identical vowel letters in a row. When the rules would give three, as in קִיֵּם kiyem "fulfilled", which would be *קייים, the writer keeps two: קיים. So the written form קיים covers kayam "exists" (adjective), kiyem "fulfilled", and kayem "fulfil!".

2.4.4 What full spelling still leaves open

Even with the rules applied, unpointed full spelling does not show:

• /a/ and most /e/ (no letter at all);
• whether ו is [v], [o] or [u], and whether י is [j], [i] or [e]/[ej];
• whether ב כ פ are stops or spirants (except final ף, ך, which are spirants in native words);
• shin vs. sin;
• gemination, and stress;
• whether an initial ו ה ב כ ל מ ש is a prefix or part of the stem.

§2.9 counts the resulting ambiguity.

2.5 The geresh: sounds that Hebrew letters lack

The geresh (גֵּרֵשׁ, ׳, U+05F3; often typed as the apostrophe ') after a letter changes its value. It marks consonants found in loanwords, slang and names, and in transcriptions of Arabic (Academy of the Hebrew Language, rules for transcribing foreign words; Wikipedia "Geresh").

Speakers have no [θ] or [ð], so ת׳ and ד׳ are pronounced [t] and [d] (or sometimes [s], [z]). They serve in transcription only.

[w]. Hebrew has no standard letter for [w]. Loanwords and names write it as a single ו at the start of a word, or as וו inside it: ווטסאפ wótsap "WhatsApp", וויסקי víski/wíski "whisky", ווינדוס wíndos "Windows". The Academy's transcription rules recommend וו at the start of a word too, and ו׳ is non-standard. In speech, [w] is often replaced by [v].

[p] and [f] at word end, and final צ׳. See §2.1.1.

2.6 Gershayim, abbreviations and acronyms

The gershayim (גֵּרְשַׁיִם, ״, U+05F4, often typed as the ASCII quotation mark ") go before the last letter of a word that is:

1. an acronym (rashei tevot, רָאשֵׁי תֵּבוֹת), made of the first letters of several words; or
2. a number written with letters of two or more letters (§2.7).

Acronyms are very common in Modern Hebrew, and many are said as words:

The geresh alone marks an abbreviation cut off after one or more letters (kitsur):

Prefixes on acronyms. A prefix attaches to the acronym, and the gershayim stay before its last letter: בצה״ל be-Tsáhal "in the IDF", לארה״ב le-Artsot ha-Brit "to the US", מהאו״ם me-ha-um "from the UN". Tokenization must not split such a word at the gershayim. In some UD treebanks, these are multiword tokens like any other word with a prefix (11. Modern Hebrew in Universal Dependencies).

Typing. In practice, ״ (U+05F4) and ׳ (U+05F3) are rare in user-generated text. Writers use ASCII " (U+0022) and ' (U+0027), and sometimes typographic quotes or ’. A normalizer should map all of these to one form, and must also handle the use of " and ' as real quotation marks (§2.8).

2.7 Numbers written with letters (gematria)

Each letter has a numerical value (table in §2.1). Numbers are written by adding the values, largest first. This "Hebrew numbering" is used for:

• Hebrew calendar dates and years: ה׳תשפ״ז = 5787;
• ordinals and section numbers in lists, laws, chapters and verses (סעיף ג׳ "section 3", פרק י״ב "chapter 12");
• school grades (כיתה ה׳ "5th grade") and days of the week (יום א׳ "Sunday");
• house numbers in some old addresses and religious texts.

Everyday arithmetic, prices and quantities use Arabic digits.

Years are usually written without the thousands: תשפ״ז = (5)787. Rule for punctuation: one letter takes a geresh (ה׳), several take gershayim before the last letter (תשפ״ז). Sources: Wikipedia "Hebrew numerals"; Academy rules for punctuation (2002).

The same convention gives rise to gematria, the interpretive practice of comparing words with equal numerical value (e.g. יין yáyin "wine" = 70 = סוד sod "secret"). It belongs to traditional exegesis and wordplay, not to grammar.

2.8 Punctuation

Hebrew uses the European set of punctuation marks: period, comma, colon, semicolon, question mark, exclamation mark, quotation marks and parentheses. The Academy's punctuation rules of 2002 (updated 2017) describe the standard use (Academy of the Hebrew Language, Klalei ha-pisuk). Points specific to Hebrew:

• Maqaf (מַקָּף, ־, U+05BE): the traditional Hebrew hyphen, set at the top of the line. It joins words in compounds and in some fixed phrases, such as בית־ספר "school", דו־שיח "dialogue" and אי־אפשר "impossible". In modern text it is often replaced by the ASCII hyphen - or by a plain space. Both spellings of the same compound must be expected.
• Quotation marks: "…" (ASCII) is the most common choice. The Academy recommends „…“ or ״…״-style marks, but these are rare in data. Because ASCII " also stands for the gershayim (§2.6), a text such as אמר "צה"ל מוכן" is ambiguous for a naive tokenizer.
• Ellipsis, dash (– or a spaced hyphen) as in English.
• Sentence-final punctuation sits at the left end of the line (the end in RTL), as the bidi algorithm places it (§2.11).
• The sof pasuq ׃ (U+05C3) and paseq ׀ (U+05C0) belong to Biblical text only.

2.9 How ambiguous is unpointed text?

Unpointed Modern Hebrew is highly ambiguous at the level of the written token. One string can be:

• different lexemes with different vowels: ספר = séfer "book", sapar "barber", safar "counted", siper "told", sfar "border region", supar "was told";
• one lexeme in different forms, or homophones across parts of speech: אהבה = ahava "love (N)" and ahava "she loved"; כתבה = katva "she wrote" and katava "(newspaper) article";
• a prefix + stem split in several ways: בצל = batsal "onion" / be-tsel "in a shadow" / ba-tsel "in the shadow"; שמן = shémen "oil" / shamen "fat (ADJ)"; מכה = maka "blow, plague" / make "hits (PRS.MS)"; הרכבת = ha-rakévet "the train" / harkavat "assembly of (CS)" / hirkavt "you (F) assembled";
• a stem + pronoun suffix: ספרה = sifra "digit" / sifra(h) "her book" / safra "she counted";
• unclear spirantization and shin/sin: בשר = basar "meat" / biser "announced" (ב = [b]/[v] in other forms of the same root); שבר = shéver "fracture" / shavar "broke".

Example of a sentence-level ambiguity with the prefix ה ha- "the":

הספר נמצא בחדר
ha-séfer nimtsa ba-khéder
DEF-book be_found.PST.3MS in.DEF-room
'The book was found / is located in the room.'

הספר נמצא בחדר
ha-sapar nimtsa ba-khéder
DEF-barber be_found.PST.3MS in.DEF-room
'The barber is in the room.'

Both readings are spelled exactly alike, and only the context decides.

2.9.1 Counted figures

From the literature on Hebrew morphological analysis:

• A study of a Modern Hebrew corpus found an average of 2.1 analyses per word form, with 55% of the word forms morphologically ambiguous (Sima'an et al. 2001, as reported by Tsarfaty 2006).
• On a 6M-word newspaper corpus (178,580 distinct word forms, 64,541 lemmas), the analyzer gave 2.4 analyses per word on average (Adler & Elhadad 2006).
• Goldberg & Elhadad (2009), citing Adler (2007): 2.7 analyses per token, compared with 1.4 in English. Some tokens have up to 15 readings.

Our own count (2026-10-17). This count uses the two CC BY-SA UD treebanks, IAHLTwiki and IAHLTknesset (UD 2.18; 129,208 non-punctuation tokens). For each surface token, we listed the distinct analyses that the gold annotation gives it, an analysis being the sequence of (lemma, UPOS) of its syntactic words. This counts only readings attested in the gold data, so it is a lower bound on the ambiguity a reader faces.

The most ambiguous types are mostly frequent function words, which is why few types cover many tokens:

(The count was made with a short script over UD_Hebrew-IAHLTwiki/*.conllu and UD_Hebrew-IAHLTknesset/*.conllu. Some "ambiguity" comes from the annotation, not from the language, such as בנוגע analysed both as one word and as two.)

Consequences for processing:

• tokens must be segmented into prefixes, stem and suffix, and the segmentation is itself ambiguous, so it has to be decided jointly with tagging and parsing (Tsarfaty 2006; Goldberg & Tsarfaty 2008; More et al. 2019);
• a lexicon of all forms with all analyses (Hspell, the MILA lexicon of Itai & Wintner 2008, Dicta's word list) is the usual first step;
• diacritization (restoring the points) is a separate, hard task in its own right (Shmidman et al. 2020, Nakdan; Gershuni & Pinter 2022).

See 11. Modern Hebrew in Universal Dependencies and 12. Mova and Modern Hebrew: where it stands, what it would need.

2.10 Prefixes written together with the word

Seven one-letter function words (traditionally mosh ve-khalev, מש״ה וכל״ב) are written attached to the following word, and can stack:

When ב, כ or ל comes before the article ה, the ה is not written and its vowel goes onto the preposition: ל + הבית → לבית la-báyit "to the house". So the unpointed לבית is both le-váyit "to a house" and la-báyit "to the house". The UD treebanks handle the hidden article in two ways. HTB restores it as a syntactic word of its own, with FORM ה_, LEMMA ה, DET and PronType=Art (למשטרה = ל + ה_ + משטרה; 2 547 such words in HTB train). IAHLTwiki and IAHLTknesset write no extra word and put Definite=Def|PronType=Art on the preposition (לבית = ל[Definite=Def] + בית). Counted in UD 2.18 on 2026-10-17. This is one of the main points in 11. Modern Hebrew in Universal Dependencies.

The order of stacked prefixes is fixed: ו, then ש/כש, then ב/כ/ל/מ, then ה.

2.11 Unicode, right-to-left layout and normalization

2.11.1 Code points

Source: The Unicode Standard, chapter on Hebrew and the code charts for U+0590–U+05FF (unicode.org/charts/PDF/U0590.pdf).

2.11.2 Normalization

• Point order. Several points can sit on one letter (dagesh + vowel + shin dot). Canonical ordering under Unicode NFC/NFD sorts combining marks by their canonical combining class. For Hebrew, these classes were assigned in a way that does not follow the traditional order of writing, and NFC can reorder points, for example putting a vowel before the dagesh. Two sources that look identical can then differ byte for byte. Compare pointed text after normalization, or compare the consonants only. (Mova's tools/macula compares lemmas by consonants for this reason.)
• Presentation forms. U+FB1D–U+FB4F decompose under NFKC/NFD into base letters + points. Apply NFD or NFKC before comparing.
• Lookalikes. Writers type ׳ ״ as ' ", the maqaf as -, and the final letters are sometimes wrong in OCR or in older encodings (e.g. ISO-8859-8 "visual" Hebrew, stored in reverse order).
• Stripping points. For matching unpointed text, remove U+0591–U+05C7 except the letters, maqaf and geresh/gershayim. Removing the points without the ktiv male changes gives a defective spelling (שֻׁלְחָן → שלחן), which differs from the everyday full spelling (שולחן). A diacritized corpus therefore cannot be turned into everyday text by simple stripping.

2.11.3 Bidirectional text

Hebrew is laid out right to left by the Unicode Bidirectional Algorithm (UAX #9). Hebrew letters are strong RTL characters. Digits and Latin letters keep their left-to-right order inside a Hebrew line. Practical consequences:

• Logical order. Text is stored in logical (reading) order and displayed reversed. Old files in "visual" order (ISO-8859-8 visual, some PDFs) must be reversed. PDF text extraction of Hebrew often yields reversed or scrambled words.
• Mixed lines. A Hebrew sentence with a number, an English word or a URL renders in a mixed order, and neutral characters (punctuation, spaces, parentheses) take their direction from their neighbours. Parentheses are mirrored: ")" is displayed as "(" in RTL.
• Control characters (U+200E LRM, U+200F RLM, U+202A–U+202E embeddings, U+2066–U+2069 isolates) are often present in web text and must be stripped or ignored by a tokenizer.
• Code blocks in this manual show Hebrew in logical order. Viewers that do not apply bidi to code blocks may show the words of a Hebrew line left to right.

2.12 Romanization standards

Several standards exist for writing Hebrew in Latin letters. They differ in what they try to show: the spelling, the Tiberian pronunciation, or modern speech.

Sources: Wikipedia "Romanization of Hebrew" (accessed 2026-10-17); Academy of the Hebrew Language, Klalei ha-ta'atik mi-ivrit le-otiyot latiniyot (2006); SBL Handbook of Style, 2nd ed. (2014), §5.1.

The Academy 2006 rules map letters as follows. Consonants: א and ע → ʾ / ʿ (or nothing), ב b/v, ג g, ד d, ה h, ו v, ז z, ח ẖ (h with a line below; plain h in simplified use), ט t, י y, כּ k, כ kh, ל l, מ m, נ n, ס s, פּ p, פ f, צ ts, ק k, ר r, ש sh, שׂ s, ת t. Vowels: a e i o u, with no length marks, and shva na written e. Unlike everyday practice, the rules keep ח (ẖ) apart from כ (kh) and do not write the diphthong [ej] of tsere-yod, but they do follow modern speech in dropping many shva na (zman, not zeman). Gemination is shown only in some cases. (Values from Wikipedia's comparison table; the Academy's own PDF should be consulted for details.)

2.12.1 The transliteration used in this manual

This manual uses a broad Israeli transliteration, meant to show how a standard speaker says the word, not how it is spelled. It is close to the Academy's 2006 rules and to SBL general style, with these choices:

• Stress is marked with an acute accent only when it is not on the final syllable: yéled "boy", báyit "house", séfer "book", khanut "shop" (final, unmarked). Final stress is the morphological default for native words, and penultimate stress is common in segolate nouns, some suffixes and loanwords. A two-syllable word without an accent is stressed on the last syllable. Monosyllables carry no accent.
• Prefixes (the article, the prepositions ב כ ל מ, ו, ש) are separated by a hyphen: ha-yéled "the boy", be-váyit "in a house", la-yéled "to the boy", ve-ha-yéled "and the boy", she-ra'iti "that I saw". This mirrors the UD segmentation into syntactic words (11. Modern Hebrew in Universal Dependencies). Pronoun suffixes are not hyphenated: sifri "my book".
• Spirantization is shown as it is pronounced: be-váyit (ב after the prefix = [v]).
• Shva is written e only where it is pronounced in standard speech (zman, but be-, le-, medabéret).
• ח and ע as pharyngeals [ħ ʕ] are not shown. Where the Mizrahi realization matters, 3. Phonology uses IPA.
• Final ה of a feminine noun is not written (yalda "girl"). The pronounced [h] of the mappiq suffix is written in parentheses: sifra(h) "her book".
• Names of people and places use their usual English spellings (Tel Aviv, Ben-Yehuda).
• IPA in square brackets [ ] is used where phonetic detail matters, and slashes / / for phonemes.

2.12.2 Example layout

Every example in this manual is a four-line block: the Hebrew as normally written (unpointed unless points matter), the transliteration, a Leipzig-style gloss, and a translation.

הילדה קראה ספר בגן
ha-yalda kar'a séfer ba-gan
DEF-girl read.PST.3FS book in.DEF-garden
'The girl read a book in the garden.'

Gloss abbreviations: 1, 2, 3 person; M masculine; F feminine; SG/S singular; PL/P plural; DEF definite; ACC accusative marker את; CS construct state; POSS possessive; PST past; PRS present (participle); FUT future; IMP imperative; INF infinitive; NEG negation; COP copula; EXIST existential; REL relative ש; Q question particle; PASS passive; DU dual. A combined person-gender-number tag is written 3MS (3rd person masculine singular), 2FP (2nd person feminine plural), and so on. Abbreviations specific to one chapter are explained there.

References

• Academy of the Hebrew Language (1947; revised; 2017). Klalei ha-ktiv she-eino menukad / Klalei ha-ktiv ha-male [Rules of spelling without points / of full spelling]. hebrew-academy.org.il/topic/hahlatot/missingvocalizationspelling (2017 PDF linked there). Automatic fetching refused (HTTP 403) on 2026-10-17; content checked via the Hebrew Wikipedia summary below.
• Academy of the Hebrew Language (2002, updated 2017). Klalei ha-pisuk [Punctuation rules]. hebrew-academy.org.il/topic/hahlatot/punctuation
• Academy of the Hebrew Language (2006). Klalei ha-ta'atik mi-ivrit le-otiyot latiniyot [Rules for transliteration from Hebrew into Latin letters]. hebrew-academy.org.il/topic/hahlatot/transliteration
• Adler, Meni & Michael Elhadad (2006). An Unsupervised Morpheme-Based HMM for Hebrew Morphological Disambiguation. COLING-ACL 2006. aclanthology.org/P06-1084 our card ; local: $RESEARCH_RAW/aclbulk-2006-adler-unsupervised-morpheme-based-hmm-hebrew/
• Asherov, Daniel & Evan-Gary Cohen (2019). A phonetic description of Modern Hebrew consonants and vowels. Brill's Annual of Afroasiatic Languages and Linguistics 11(1). (cited via Wikipedia; not held locally)
• Bolozky, Shmuel (1997). Israeli Hebrew phonology. In A. S. Kaye (ed.), Phonologies of Asia and Africa, vol. 1, 287–311. Eisenbrauns. (not held locally)
• Coffin, Edna Amir & Shmuel Bolozky (2005). A Reference Grammar of Modern Hebrew. Cambridge University Press. (not held locally; listed in $RESEARCH_RAW/hebgram/grammars.tsv)
• Gershuni, Elazar & Yuval Pinter (2022). Restoring Hebrew Diacritics Without a Dictionary. Findings of NAACL 2022. aclanthology.org/2022.findings-naacl.75 our card ; local: $RESEARCH_RAW/findings-2022-gershuni-restoring-hebrew-diacritics-without-dictionary/
• Goldberg, Yoav & Michael Elhadad (2009). Hebrew Dependency Parsing: Initial Results. IWPT 2009. aclanthology.org/W09-3819 our card ; local: $RESEARCH_RAW/lingbulk-2009-goldberg-hebrew-dependency-parsing-initial-results/
• Goldberg, Yoav & Reut Tsarfaty (2008). A Single Generative Model for Joint Morphological Segmentation and Syntactic Parsing. ACL 2008. aclanthology.org/P08-1043
• Har'El, Nadav & Dan Kenigsberg. Hspell — the free Hebrew spell checker and morphological analyzer, v1.4. hspell.ivrix.org.il ; local: $RESEARCH_RAW/lexres/morphology/sa/hspell/ (AGPL-3.0)
• Hebrew Wikipedia. כתיב חסר ניקוד [Spelling without points]. he.wikipedia.org/wiki/כתיב חסר ניקוד (accessed 2026-10-17)
• Itai, Alon & Shuly Wintner (2008). Language resources for Hebrew. Language Resources and Evaluation 42(1), 75–98. (see also $RESEARCH_RAW/lingbulk-2006-itai-computational-lexicon-contemporary-hebrew/)
• Khan, Geoffrey (2020). The Tiberian Pronunciation Tradition of Biblical Hebrew, 2 vols. Open Book Publishers. doi.org/10.11647/OBP.0163 our card ; local: $RESEARCH_RAW/hebgram/open/khan-tiberian-pronunciation-v1-2020/
• Laufer, Asher (1999). Hebrew. In Handbook of the International Phonetic Association, 96–99. Cambridge University Press.
• More, Amir, Amit Seker, Victoria Basmova & Reut Tsarfaty (2019). Joint Transition-Based Models for Morpho-Syntactic Parsing: Parsing Strategies for MRLs and a Case Study from Modern Hebrew. TACL 7. aclanthology.org/Q19-1003 our card
• SBL (2014). The SBL Handbook of Style, 2nd ed., §5.1 (Hebrew transliteration). SBL Press.
• Shmidman, Avi, Shaltiel Shmidman, Moshe Koppel & Yoav Goldberg (2020). Nakdan: Professional Hebrew Diacritizer. ACL 2020 System Demonstrations. aclanthology.org/2020.acl-demos.23 our card ; local: $RESEARCH_RAW/aclbulk-2020-shmidman-nakdan-professional-hebrew-diacritizer/
• Sima'an, Khalil, Alon Itai, Yoad Winter, Alon Altman & Noa Nativ (2001). Building a tree-bank of Modern Hebrew text. Traitement Automatique des Langues 42(2). (cited via Tsarfaty 2006)
• Tsarfaty, Reut (2006). Integrated Morphological and Syntactic Disambiguation for Modern Hebrew. COLING-ACL 2006 Student Research Workshop. aclanthology.org/P06-3009 our card ; local: $RESEARCH_RAW/aclbulk-2006-tsarfaty-integrated-morphological-syntactic-disambiguation-modern/
• Unicode Consortium. The Unicode Standard, Hebrew block U+0590–U+05FF code chart. unicode.org/charts/PDF/U0590.pdf ; and UAX #9, Unicode Bidirectional Algorithm. unicode.org/reports/tr9
• Universal Dependencies 2.18: UD_Hebrew-IAHLTwiki and UD_Hebrew-IAHLTknesset (CC BY-SA 4.0). universaldependencies.org/treebanks/he iahltwiki ; local: $RESEARCH_RAW/lexres/treebanks/ud-2.18/
• Wikibooks (Hebrew). Dikduk ha-Ivrit ha-Hadasha [Grammar of Modern Hebrew], pages "Consonants" and "Vowels" (CC BY-SA). he.wikibooks.org/wiki/דקדוק העברית החדשה ; local: $RESEARCH_RAW/hebgram/sa/wikibooks-he-dikduk-haivrit-hahadasha/
• Wikipedia. Hebrew alphabet; Geresh; Gershayim; Hebrew numerals; Romanization of Hebrew; Modern Hebrew phonology. en.wikipedia.org/wiki/Hebrew alphabet etc. (accessed 2026-10-17)

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/he/02-script-and-spelling/