Tamil to English Transliteration

If you came here to type Tamil, read this first

Most people searching for "English to Tamil transliteration" want a Tamil typing tool: you type vanakkam on an ordinary keyboard and you get வணக்கம். That is the opposite direction from what this page does, and we have not shipped it, because doing it badly is easy and doing it well is not.

The difficulty is that Tamil distinguishes sounds that English spelling does not. A word typed in Latin letters usually has several valid readings, and only one of them is the word you meant:

A straight letter-mapping table cannot choose between those columns, because the information needed to choose is not in what you typed. We tested this rather than assuming it: running ten everyday Tamil words through the two best-known open phonetic schemes, Azhagi and Jaffna, as published in the MIT-licensed Open-Tamil library, produced three correct words out of ten with Azhagi and two out of ten with Jaffna. Both schemes turned tamil into டமில் and vanakkam into வனக்கம்.

The tools that do this well, such as Google's input method and the open-source GoVarnam, do not rely on the table alone. They generate every candidate the table allows and then pick between them with a dictionary of real Tamil words, often offering you a list to choose from. That is the version worth building, and it is what we are working towards.

What the Tamil to Latin output looks like

The tool follows ISO 15919, which gives one Latin letter or letter-plus-diacritic for each Tamil letter. The result reads a little differently from the spellings you see on signs and passports, because those follow English convention rather than a standard:

Neither column is wrong. The standard column is reversible: from ceṉṉai you can reconstruct the exact Tamil spelling, which is the whole point of a transliteration standard. The apostrophe in am'mā is part of that: it marks a doubled consonant, so the result cannot be confused with a different Tamil spelling that would otherwise look the same. The conventional column is easier to read aloud but loses information, which is why two different Tamil words can end up with the same English spelling.

The letters that need diacritics

Tamil has three l-sounds, three n-sounds and two r-sounds where English has one of each. The dots and lines under the Latin letters are not ornament: they are the only thing keeping those letters apart.

ழ is the letter Tamil speakers point to when they say the language cannot be written properly in English letters. It is a retroflex approximant, close to nothing in English, and it is in the name of the language itself: தமிழ், tamiḻ.

Where the conversion data comes from

We are not linguists and we do not invent romanization rules. Tamil conversion runs on the Unicode Common Locale Data Repository transliteration data through PHP's Transliterator class, the same ICU implementation used by operating systems and browsers. The rules follow ISO 15919, maintained by the International Organization for Standardization. Where our output differs from raw ICU we say so: for Indic scripts we remove the apostrophes ICU inserts and apply schwa deletion, which is documented on the About page.

What this tool does not do

• It does not translate. வணக்கம் comes back as vaṇakkam, which is how it sounds. What it means is "greetings", and that is a translator's job.
• It does not type Tamil for you. English to Tamil needs a dictionary-backed input method, not a conversion table, for the reasons set out above.
• It does not produce passport spellings. ISO 15919 output keeps the diacritics. Official documents usually want the conventional English form instead.
• It does not handle Tamil written in other scripts. Tamil in Arabic script (Arwi) or in Brahmi is outside what the Unicode data covers.