Romanization of Hebrew
For example, the Hebrew name spelled יִשְׂרָאֵל ("Israel") in the Hebrew alphabet can be romanized as Yisrael or Yiśrāʼēl in the Latin alphabet.
Romanization includes any use of the Latin alphabet to transliterate Hebrew words. Usually it is to identify a Hebrew word in a non-Hebrew language that uses the Latin alphabet, such as German, Spanish, Turkish, and so on. The term transliteration means using an alphabet to represent the letters and sounds of a word spelled in another alphabet, whereas the term transcription means using an alphabet to represent the sounds only. Romanization can do both.
To go the other way, that is from English to Hebrew, see Hebraization of English. Both Hebraization of English and Romanization of Hebrew are forms of transliteration. Where these are formalized these are known as "transliteration systems," and, where only some words, not all, are transliterated, this is known as "transliteration policy."
When to transliterate 
Transliteration assumes two different script systems. The use of a French word in English without translation, such as "bourgeois" is not transliteration. The use of a Hindi word in English such as "khaki" (originally खाकी) is transliteration. Transliteration of a foreign word into another language is usually the exception to translation, and often occurs when there is something distinctive about the word in the original language, such as a double entendre, uniqueness, religious, cultural or political significance. Or it may occur to add local flavour.
In the cases of Hebrew transliteration into English many Hebrew words have a long history of transliteration, for example Amen, Satan, ephod, Urim and Thummim have traditionally been transliterated, not translated. These terms were in many cases also first transliterated into Greek and Latin before English.
Different publishers have different transliteration policies. For example ArtScroll publications generally transliterate more words relative to sources such as the Jewish Encyclopedia 1911, or Jewish Publication Society texts.
How to transliterate 
There are various transliteration standards or systems for Hebrew-to-English, no one system has significant common usage across all fields. Consequently in general usage there are often no hard and fast rules in Hebrew-to-English transliteration, and many transliterations are an approximation due to lack of equivalence between the English and Hebrew alphabets. Conflicting systems of transliteration often appear in the same text, as certain Hebrew words tend to associate with certain traditions of transliteration. For example,
- "For Hanukkah at the synagogue Beith Sheer Chayyim, Isaac donned his talis that Yitzchak sent him from Bet Qehila in Tsfat, Israel."
With numerous transliteration systems in use, the inconsistency is wild.
Historic instances 
Early romanization of Hebrew occurred with the contact between the Romans and the Jews. It was influenced by earlier transliteration into the Greek language. For example, the name of the Roman province of Iudaea (63 BCE) was apparently derived from the Greek words Ἰούδα (Iouda) and Ἰουδαία (Ioudaia). These words can be seen in Chapter 1 of Esdras (Ezra) in the Septuagint, a Hellenistic translation of the Hebrew Bible into Greek. The Greek words in turn are transliterations of the Hebrew word יהודה (Y'huda) that we now know adapted in English as the names Judah, Judas and Jude.
In the 1st century, Satire 14 of Juvenal uses the Hebraic words sabbata, Iudaicum, and Moyses, apparently adopted from the Greek.
The 4th century and 5th century Latin translations of the Hebrew Bible romanize its proper names. The familiar Biblical names in English are derived from these romanizations. The Vulgate, of the early 5th century, is considered the first direct Latin translation of the Hebrew Bible. Apart from names, another term that the Vulgate romanizes is the technical term mamzer (Hebrew ממזר).
With the rise of Zionism, some Jews promoted the use of romanization instead of Hebrew script in hopes of helping more people learn Hebrew. One such promoter was Ithamar Ben Yehuda, or Ittamar Ben Avi as he styled himself. His father Eliezer Ben Yehuda raised him to be the first modern native speaker of Hebrew. In 1927 Ben-Avi published the biography Avi in romanized Hebrew (now listed in the online catalog of the Jewish National and University Library). However, the innovation did not catch on.
Modern uses 
Romanized Hebrew can be used to present Hebrew terminology or text to anyone who is not familiar with the Hebrew script. Many Jewish prayer books include supplementary romanization for some or all of the Hebrew-language congregational prayers.
Romanized Hebrew is also used for Hebrew-language items in library catalogs and Hebrew-language place names on maps. In Israel, most catalogs and maps use the Hebrew script, but romanized maps are easily available and road signs include romanized names. Some Hebrew speakers use romanization to communicate when using internet systems that have poor support for the Hebrew alphabet.
Standard romanizations exist for these various purposes. However, non-standard romanization is widely seen, even on some Israeli street signs. The standards are not generally taught outside of their specific organizations and disciplines.
- Traditional, scholarly: ISO 259:1984; ISO 259-2:1994 (simplified); Society of Biblical Literature (SBL) Handbook of Style
- National: Rules of Transcription: Romanization of Hebrew. Academy of the Hebrew Language, 1957. Updated and augmented with a simplified version, 2000. Replaced by a new system altogether in 2006.
- Bibliographic data: ANSI Z39.25-1975; ALA/LC Romanization Tables (1991) and their book Hebraica Cataloging (1987), with Encyclopaedia Judaica (1972–1993) as an authority on names and common terms. Library of Congress Authorities is an online database that records and sources the forms of subjects, names, and titles that the Library of Congress uses.
- Geographic names: BGN/PCGN 1962 (US and UK), approximately equivalent to UNGEGN 1977 (United Nations), as both are based on the Academy of the Hebrew Language recommendations. However, BGN provides more and somewhat different specific recommendations. The GEONet Names Server is an authoritative online database that lists BGN names and assists with font character availability and conventional forms of names.
- Phonemic: ISO/FDIS 259-3:1999 (not an adopted standard)
Comparative table 
The following table is a breakdown of each letter in the Hebrew alphabet, describing its name or names, and its Latin script transliteration values used in academic work. If two glyphs are shown for a consonant, then the left-most glyph is the Final form of the letter (or right-most glyph if your browser doesn't support right-to-left text layout). The conventions here are ISO 259, the UNGEGN system based on the old-fashioned Hebrew Academy system, and the modern common informal Israeli transcription. In addition, an International Phonetic Alphabet pronunciation is indicated—historical (Tiberian vocalization) for ISO 259, prescribed for Hebrew Academy, and in practice for Israeli. For the vowels further down, the letters ח and ט are used as symbolic anchors for vowel symbols, but should otherwise be ignored.
For the letters בּ גּ דּ כּ פּ תּ with dagesh in ISO 259 Classical Hebrew and by the Hebrew Academy standard, they are transcribed as single graphemes (b g d k p t) at the beginnings of words, after other consonants, and after shewa ְ or ẖatafim ֱ ֲ ֳ . In almost every other situation, they are transcribed as double letters (bb gg dd kk pp tt). This does not apply to common Israeli Hebrew transliteration, where there are no double consonants.
The letters א ה at the ends of words without additional niqqud are silent and not transliterated. The letter ו at the end of a word with ẖolam ֹ is also silent and not transliterated. The letter י at the end of a word after ẖiriq ִ is also silent and not transliterated. The situation of the letter י at the end of a word after ẕere ֵ or seggol ֶ is more complicated, as they are silent in Classical Hebrew and in Hebrew Academy prescription and not transliterated in those systems, but they form diphthongs (ei) in Israeli Hebrew—see the vowels and diphthongs sections further down. In any event, the shewa naẖ is placed between two adjacent consonants in all situations; if there is not even a shewa naẖ between consonants, then the first of the two consonants is silent and not transliterated—this is usually one of א ו י, but even occasionally ה and rarely ש (in the name יִשָּׂשכָר Issachar) are encountered silent in this fashion. In Israeli Hebrew transcription, a vowel before yud at the end of a word or before yud then shewa naẖ inside a word, is transcribed as a diphthong (ai oi ui)—see the diphthongs section further down.
In Classical Hebrew transliteration, vowels can be long (gāḏōl), short (qāṭān) or ultra short (ḥăṭep̄), and are transliterated as such. Ultra short vowels are always one of šəwā nāʻ ְ , ḥăṭep̄ səḡōl ֱ , ḥăṭep̄ páṯaḥ ֲ or ḥăṭep̄ qāmeṣ ֳ . Šəwā ְ is always always šəwā nāʻ (pronounced) if it's immediately after the word's first consonant, or after a consonant after a long vowel and before another pronounced consonant—otherwise, šəwā is realized as šəwā nāḥ (silent). The vowels ṣērē ֵ and ḥōlem ֹ are always long in all situations. The vowels ḥīreq ִ , səḡōl ֶ , páṯaḥ ַ , qāmeṣ ָ , qibbūṣ ֻ and šūreq וּ are always long if they are the stressed syllable, or if they are in a syllable before only one consonant and another vowel, and in these cases they are transliterated as long. If they are unstressed and before a double consonant or a consonant cluster, or in the word's final unstressed syllable, then they are always short and transliterated as short. But if a vowel carries an accent ֫ or a meteg ֽ , then it is always long—a meteg in particular is often used in places where a vowel is long but not necessarily the word's stressed syllable. Lastly, there are exceptional circumstances when long vowels—even ṣērē and ḥōlem—may not force a following šəwā to become šəwā nāʻ, including for example names such as גֵּרְשֹׁם Gēršōm (not Gērəšōm as it might seem), בֵּלְשַׁאצַּר Bēlšaṣṣạr (not Bēləšaṣṣạr) and צִֽקְלַג Ṣīqlạḡ (not Ṣīqəlạḡ). Some of these seem to be learned exceptions, and most words under the same circumstances have šəwā nāʻ as expected, such as נָֽצְרַת Nāṣərạṯ (not Nāṣrạṯ). (This is all moot in Israeli Hebrew, where, as already mentioned, shva nach tends to opportunistically replace shva na where comfortable, so נָֽצְרַת is Natzrat not Natzerat, etc.)
For the vowel qamaẕ ָ, whether the vowel is long or short in Classical Hebrew affects the pronunciation in Academy or Israeli Hebrew, even though vowel length is not phonemic in those systems, and the difference is transliterated accordingly. Qamaẕ qatan when short is /o/, except when at the end of a word when not before a final consonant, in which case it is /a/. Qamaẕ gadol is usually /a/, but in rare situations in Classical Hebrew it can be treated as a long open /ọ/, which although pronounced identically to /ā/ (both were [ɔː]), this a/o distinction is clearly made in the pronunciation of Academy and Israeli Hebrew, and is thus transliterated.
If any word ends with one of הַּ חַ עַ, then the vowel pataẖ is pronounced before the consonant, not after as it is written, and so the transliterated sequence is /ah/, /aẖ/, /aʻ/, etc.
In certain rare words that are meant to begin with two consecutive consonants even in Classical Hebrew, an invisible səḡōl qāṭān vowel is pronounced before the two consonants in Classical Hebrew and is so transcribed, because Classical words may not begin with more than one consonant. This rule does not apply to Academy and Israeli Hebrew, where consonant clusters are more tolerated. For example, the word שְׁתַּ֫יִם ("two") would appear as štáyim, but is actually ʼeštáyim. However, it remains simply shtayim in Academy and Israeli Hebrew.
In 2006, the Hebrew Academy replaced their 1953 transliteration rules with new rules, and these were adopted as a United Nations standard in 2007. As of 2008[update], migration to the new transliteration standard is still underway, and many signs and documents still use the 1953 conventions. The new 2006 rules attempt to more closely follow Israeli Hebrew vowel habits (such as the collapse of many shva na), but stop short of adopting most of the informal transliteration patterns. It still transliterates the diphthong [e̞͡ɪ] as /e/, and it still transliterates separate /ẖ/ and /kh/ in all cases. It is unspecific about rules governing the transliteration of phonemes not traditionally native to Hebrew.
|Symbol||Common Israeli||Hebrew Academy||ISO 259|
|א||alef||'[note 1]||-||alef||'[note 1]||alef||ʼ[note 2]||ʾālep̄||ʾ||[ʔ]|
|ו||vav||v (w)[note 5]||[v] ([w])[note 5]||vav||v||waw||w||wāw||w||[v] [(w)][note 5]|
|וּ[note 6]||vv||ww||wāw ḥāzāq||ww||[vv]|
|ח||chet||ch, kh, h||[χ]||ẖet||ẖ||ẖet||ẖ||ḥēṯ||ḥ||[ħ]|
|י||yud||y, i[note 7]||[j]||yud||y||yud||y||yōḏ||y||[j]|
|ך כ||chaf||ch, kh||[χ]||khaf||kh||khaf||kh||ḵāp̄||ḵ||[x]|
|ע||ayin||'[note 1]||-||ayin||'[note 1]||ʻayin||ʻ||ʿáyin||ʿ||[ʕ]|
|ףּ פּ[note 8]||pei||p||[p]||pe||p||pe||p||pē||p||[p]|
|ץ׳ צ׳[note 3]||tshadi||tsh, ch||[t͡ʃ]||čadi||č|
|רּ[note 4]||rr||rr||rēš ḥāzāq||rr||[ʀʀ]|
|Forms used only in transliterations of Arabic|
|ט׳[note 4]||ẓāʾ||ẓ||[ðˤ] ~ [zˤ]|
|ע׳ ר׳[note 4]||ġayn||ġ||[ɣ] ~ [ʁ]|
|ץ׳ צ׳[note 4]||ḍād||ḍ||[dˤ]|
|טְ||shva nach||shva naẖ||shewa naẖ||šəwā nāḥ|
|shva na||e[note 9]||[e̞][note 9]||shva na||e[note 9]||shewa naʻ||e||šəwā nāʻ||ə||[ɐ̆]
|חֱ||chataf segol||e||[e̞]||ẖataf seggol||e||ẖataf seggol||e||ḥăṭep̄ səḡōl||ĕ||[ɛ̆]|
|חֲ||chataf patach||a||[ä]||ẖataf pataẖ||a||ẖataf pataẖ||a||ḥăṭep̄ páṯaḥ||ă||[ɐ̆]|
|חֳ||chataf kamatz||o||[o̞]||ẖataf kamats||o||ẖataf qamaẕ||o||ḥăṭep̄ qāmeṣ||ŏ||[ɔ̆]|
|səḡōl gāḏōl||é, ẹ||[ɛː]|
|páṯaḥ gāḏōl||á, ạ||[ɐː]|
|טָ||kamatz gadol||kamats gadol||qamaẕ gadol||qāmeṣ gāḏôl||ā||[ɔː]|
|kamatz katan||o||[o̞]||kamats katan||o||qamaẕ qatan||o||ọ|
|טוּ[note 6]||shuruk||shuruk||shuruq||šūreq qāṭān||u||[u]|
|טֶי||segol yud||seggol||seggol||səḡōl||e, é, ẹ||[ɛ(ː)]|
|טַי טַיְ||patach yud||ai||[ä͡ɪ]||pataẖ yud||ay||pataẖ yud||ay||páṯaḥ yōḏ||ay, áy, ạy||[ɐ(ː)j]|
|טָי טָיְ||kamatz gadol yud||kamats gadol yud||qamaẕ gadol yud||qāmeṣ yōḏ||āy, oy, ọy||[ɔ(ː)j]|
|kamatz katan yud||oi||[o̞͡ɪ]||kamats katan yud||oy||qamaẕ qatan yud||oy|
|טֹי טֹיְ||cholam yud||ẖolam yud||ẖolam yud||ḥōlem yōḏ||ōy||[oːj]|
|טֻי טֻיְ||kubutz yud||ui||[u͡ɪ]||kubbuts yud||uy||qubbuẕ yud||uy||ḥōlem yōḏ||uy, ūy||[u(ː)j]|
|טוּי טוּיְ||shuruk yud||shuruk yud||shuruq yud||šūreq yōḏ|
- Functions only as a syllable break in Israeli Hebrew. So, when transcribed, is omitted everywhere except in circumstances where omission would be ambiguous, such as immediately after consonants or between vowels.
- Omitted at the beginnings of words.
- For phonemes in loanwords or their derivations. These are not used in Classical Hebrew, so are omitted from the ISO 259 columns.
- Hebrew-letter transliterations of foreign phonemes (used only for the transliterition of proper names or foreign language texts, never in native words or loanwords).
- Rarely, loanwords or their derivations contain the phoneme /w/, which standardly is spelled with "ו", indistinguishably from "ו" denoting /v/. Some non-standard distinguishing ways of spelling exist, such as "ו׳" or word-initial "וו" (see orthographic variants of vav), but their usage is not consistent. Therefore, correctly romanizing the "ו" as /w/ cannot rely on orthography but rather on lexical knowledge, e.g. "דַּאַוִין" Morfix → "da'awín", "בָּאֲוִיר" Morfix → "ba'avír".
- Shuruk and "ו" with a dagesh look identical ("וּ") and are only distinguishable through the fact that in text with nikud, "ו" with a dagesh will normally be attributed a vocal point in addition, e.g. שׁוּק, ("a market"), trans. "shuk" (the "וּ" denotes a shuruk) as opposed to שִׁוֵּק ("to market"), trans. "shivék" (the "וּ" denotes a vav with dagesh). In the word שִׁוּוּק ("marketing"), trans. "shivúk", the first ("וּ") denotes a vav with dagesh, the second a shuruk., being the vocal point attributed to the first.
- Usually written as i after a vowel and either before a consonant or syllable break or at the end of a word.
- The pe hazak sofit ףּ can be found sometimes in the Tanakh. But ף is never used for /p/ in Modern Hebrew, with פ being used in all positions, including word-finally.
- Israeli Hebrew typically demotes shva na to silent shva nach in situations where it can be comfortably omitted in common speech, creating consonant clusters that would otherwise not be permitted in older varieties of Hebrew. In situations where this happens, the shva is typically not transliterated at all, or is at most transliterated with an apostrophe (') – see shva.
Transcription vs. transliteration 
Different purposes call for different choices of romanization. One extreme is to make a phonetic transcription of one person's speech on one occasion.
In Israel, a pronunciation known as General Israeli Hebrew or Standard Hebrew is widely used and documented. For Israeli speech and text where linguistic groups are not at issue, romanization can use a phonetic transcription according to Standard Hebrew pronunciation. However, there are many Israeli groups with differing pronunciations of Hebrew and differing social priorities.
An attempt to devise a more general system of romanization is complicated by the long and varied history of the Hebrew language. Most Hebrew texts can be appropriately pronounced according to several different systems of pronunciation, both traditional and modern. Even today, it is customary to write Hebrew using only consonants and matres lectionis. There was no way to indicate vowels clearly in Hebrew writing until the time of the Second Temple. Since an earlier time, multiple geographically separated communities have used Hebrew as a language of literature rather than conversation.
One system of assigning and indicating pronunciation in Hebrew, the Tiberian vocalization, is broadly authoritative for Hebrew text since the end of the Second Temple period (Sáenz-Badillos, page xi). It is possible to accommodate the pronunciations of different communities by transliterating the Tiberian vocalization without attempting to transcribe a specific phonetic pronunciation.
Notable varieties of Hebrew for which Tiberian vocalization is not suitable are the Hebrew of the Qumran community (as known from the Dead Sea Scrolls) and of the Samaritans. For romanizations of Samaritan pronunciation, it is advisable to take quotations directly from a Samaritan edition of the Hebrew Bible, which has approximately 6,000 textual variations from Jewish editions.
It is appropriate to focus only on the consonantal spelling when discussing unusually structured words from ancient or medieval works.
Use of Tiberian principles 
The Tiberian vocalization was devised in order to add indications of pronunciation to the consonantal text of the Hebrew Bible, without changing the consonantal text. It was intended for experts in Biblical Hebrew grammar and morphology.
Transliterations usually avoid the typographically complex marks that are used in Tiberian vocalization. They also attempt to indicate vowels and syllables more explicitly than Tiberian vocalization does. Therefore a technical transliteration requires the use of Tiberian principles, as mentioned below, rather than simply representing the Tiberian symbols. Many transliteration standards require a thorough knowledge of these principles, yet they usually do not provide practical details.
- There are seven basic vowels.
- A vowel may be long, short, or ultrashort.
- The vowel "shva" may be sounded (shva na) or silent (shva nach).
- Consonants that have been used historically to indicate vowels, the "matres lectionis", are no substitute for proper vowel marks.
- The vowel "kamets" may have its usual sound (kamets gadol - long "a") or a different sound (kamets katan - short "o").
- Six consonants (beth, gimel, daleth, kaph, pe, and tav) can be hard or soft. To be specific, they are pronounced either as stops or fricatives ("spirantized"). For example, the letter bet can be pronounced as "b" or "v". Tiberian vocalization marks a hard consonant with a dagesh kal (in the Hebrew term) or lene (Latin). A soft consonant lacks a dagesh kal, and is sometimes explicitly marked using rafe, an overbar. Transliterations sometimes also use an overbar or underbar to mark a soft consonant. (In Modern Sephardic Hebrew, however, only three consonants—bet, kaph, and pe—retain the hard-soft distinction. In the Ashkenazic style of pronunciation, the soft tav is sounded as "s".)
- A letter that looks like shin may be that letter (when marked with a shin dot) or the letter sin (when marked with a sin dot).
- Most consonants can undergo gemination. Tiberian vocalization marks gemination with a dagesh hazak (in the Hebrew term) or forte (Latin), which looks the same as dagesh kal.
- A consonant that is normally silent (most often he) may be sounded if it is a root consonant or possessive ending. Tiberian vocalization marks such a consonant using a mapiq, which looks like a dagesh.
- A silent vav may be used to hold a holem vowel, but sometimes a vav with holem has consonant value.
Additional transliteration principles 
A further complication is that the Roman alphabet does not have as many letters for certain sounds found in the Hebrew alphabet, and sometimes no letter at all. Some romanizations resolve this problem using additional non-Tiberian principles:
- The two letters that represent a stop may be written using the forward and backward quote marks, or similar marks.
- Certain consonants are considered "emphatic" (the consonants ח ט צ), due to being pronounced traditionally toward the back of the mouth. They may be transliterated distinctively by using an underdot.
- The letter "vav" (ו) was once pronounced like English "w", in contrast to its current pronunciation identical to the letter "vet" (the soft letter ב).
- The Karmeli transcription (see link at bottom of page) creates additional letters based on similar Hebrew or Cyrillic letters to represent the sounds which lack Roman letters.
Finally, for ease of reading it is common to apply certain principles foreign to Hebrew:
- Use a hyphen between common prefixes or suffixes and a romanized word.
- Capitalize the first letter of a proper name, but not its prefixes.
See also 
- Hebraization of English
- Help:IPA for Hebrew
- Arabic transliteration
- Other alphabets in Morse code#Hebrew
- e.g. rationale of the author of Athenian settlements of the fourth century B.C. p.xiii Jack Cargill - 1995 "In relation to Greek words other than personal names, the main issue is deciding when to transliterate, ie, when to employ an Anglicized Greek word rather than translating the Greek word into English. My tendency is "when in doubt, ..."
- Exodus p.152 Henry Leopold Ellison - 1982 "THE PRIESTLY GARMENTS II Exodus 28:1-43 (cont'd) THE EPHOD (28:5 14) We have no means of giving a definitive meaning to "ephod", the English being simply a transliteration of the Hebrew."
- REPORT ON THE CURRENT STATUS OF UNITED NATIONS ROMANIZATION SYSTEMS FOR GEOGRAPHICAL NAMES, Compiled by the UNGEGN Working Group on Romanization Systems Version 2.2, January 2003
- REPORT ON THE CURRENT STATUS OF UNITED NATIONS ROMANIZATION SYSTEMS FOR GEOGRAPHICAL NAMES, Compiled by the UNGEGN Working Group on Romanization Systems Version 3.0a, May 2008
- Announcements of the Academy of the Hebrew Language
- Angel Sáenz-Badillos, A History of the Hebrew Language. Cambridge University Press, 1993. Reprinted in paperback 2000, ISBN 0-521-55634-1.
- "The Jewish diaspora: Rome" at livius.org. Retrieved August 9, 2005.
- Parallel texts of Ezra in various languages at Unbound Bible. Retrieved August 16, 2005.
- Juvenal, Satire 14 at The Latin Library. Retrieved August 9, 2005.
- "Transliteration" at nostradamus.net. Retrieved August 9, 2005. Excellent, lively summary of issues and options for transliteration of Hebrew.
- The entire Hebrew bible transliterated in two different formats, Matthew Anstey
- Transliteration between Hebrew and Standard English or ISO 259
- "Hebrish or Engbrew", Talya Halkin in Jerusalem Post (July 6, 2006).
- "Rules of Transcription from Hebrew Script to Latin Script" (in Hebrew) of the Academy of the Hebrew Language, 2007.
- United Nations romanization of Hebrew
- GEONet Names Server
- ALA/LC romanization of Hebrew
- Library of Congress Authorities
- Society of Biblical Literature, SBL Handbook of Style.
- Transliterated text of the Torah and all Haftarot are available from http://bible.ort.org/ based on Sephardi pronunciation.
- The Karmeli Alphabet; a Romanization developed in Israel which uses extra invented letters
- Online Hebrew Transliterator, supports 40 languages
- Automatic Hebrew-To-English Transliteration