Random Language Generator

How this random language generator works

Press Generate and you get a language and its ISO code. Turn up the Number box for several at once, no repeats. Copy takes them.

English (en). Aghem (agq). Sumerian (sux). Klingon (tlh).

Look at the codes. Two letters, then three. That difference is not a formatting choice, and it is the most interesting thing about the way the world catalogues its own speech.

A hundred and eighty-four, and no more, ever

ISO 639-1, the two-letter standard, contains exactly 184 codes. Not approximately. Exactly. That is every two-letter language code that exists in the world.

Two letters gives you 676 possible combinations, and some of those are unusable. Meanwhile the world has roughly seven thousand living languages, and ISO 639-3 registers close to 7,900 codes once you count the extinct, the ancient and the constructed.

The two-letter space was never going to be enough. It went to what ISO calls the major, mostly national languages: the ones with tens of millions of speakers, official status somewhere, a literature, an interest from institutions.

So en is English and fr is French, and Aghem, spoken in Cameroon, is agq. Not because Aghem is a lesser language. Because the alphabet ran out, and English was there first.

Nothing new will ever get two letters. The list is closed. Whatever language is spoken in your street, if it did not have an empire behind it in the twentieth century, it has three letters and it always will.

A language is not a locale

Software people confuse these constantly, and it causes real bugs.

A language is a language. en. A locale is a language plus a place plus a set of conventions about dates, currency, sorting and decimal points. en_GB. fr_CH. es_419, which is Latin American Spanish, where 419 is the UN region code for Latin America.

American English and British English are the same language having a disagreement about the letter u. They are two locales and one language, and a system that treats them as two languages will translate your interface twice for no reason.

Unicode maintains the definitive locale data, and it contains one entry that gives the whole game away. There is a locale called root. It is not a language. It is the fallback: the setting a system lands on when it does not know what language it is dealing with. It is a shrug, given a name and a code.

No language has ever been called Root. But every piece of software you own has spoken it at least once.

Chinese is not a language. It is a container.

Ask for the code for Chinese and you get zh. Then look at what sits underneath it.

Mandarin, Wu, Yue (which most English speakers call Cantonese), Min Nan, Hakka, Gan, Xiang. Each has its own ISO 639-3 code. Several are mutually unintelligible in speech. A Cantonese speaker and a Mandarin speaker, talking, understand each other about as well as a Portuguese speaker and an Italian.

ISO calls zh a macrolanguage: a code that covers a group of closely related languages which, in some contexts, are treated as one. Arabic is another. So is Malay, and Quechua, and Zapotec.

The category exists because the question "is this a language or a dialect?" has no linguistic answer. Mutual intelligibility is a spectrum, not a line, and where the line gets drawn is decided by armies, borders, literatures and pride. The old joke is that a language is a dialect with a navy.

ISO's response was to refuse to adjudicate, and to give both answers a code.

Languages nobody speaks, and one nobody ever did

ISO 639-3 does not restrict itself to the living. Its remit covers languages living, extinct, ancient, historical and artificial.

The dead are in it. Latin (la). Sumerian (sux), which stopped being spoken in antiquity. Akkadian (akk). Hittite (hit). Gothic (got). Old English (ang), which is a foreign language to every English speaker alive.

The invented are in it too. Esperanto (eo), which has a two-letter code and therefore outranks Aghem in the eyes of the standard. Volapük (vo). Ido (io). Lojban (jbo). Afrihili (afh), a constructed auxiliary language.

And Klingon (tlh).

Klingon was written for a television programme. It has a registered three-letter ISO code, which places it, administratively, on exactly the same footing as Sumerian and Hittite and several hundred languages spoken by living people in the mountains of Papua.

That is not a mistake in the standard. It is the standard doing its job. A code is an identifier, not a compliment.

Ways people actually use this

• Test data for localisation. The codes are correct and the names are the official English ones. That is the whole job.
• Quizzes. Show the code, ask the language. Then show tlh and watch what happens.
• Fiction. Your character speaks four languages. Generate four and see whether the combination tells you where they grew up.
• Perspective. Generate twenty. Count how many you have heard of. The answer is usually two or three, out of about seven thousand.
• Not a speaker count. Nothing here tells you how many people speak a language. Chinese and Zenaga are weighted exactly the same by the sampler.

Getting more out of the filters

• Contains matches anywhere, including inside the code. Type (a or (z to slice by first letter of code rather than of name.
• Type Old, Ancient, Middle or Classical to find the languages that stopped.
• Type Chinese and read the results carefully. One is the macrolanguage. Most of the rest are its members. One, Achinese, is spoken in Sumatra and has nothing to do with China, because the filter ignores case.
• Type Creole or Sign. Both are whole families of language that most people never think to look for.
• Starts with and Ends with compare a single character, so give them one letter.

Questions people ask

How many languages are there?

Roughly seven thousand living ones. ISO 639-3, which also covers extinct, ancient and constructed languages, registers close to 7,900 codes.

Why do some languages have two-letter codes and others three?

ISO 639-1 assigns two letters, and it contains only 184 codes, reserved for major and mostly national languages. Everything else is identified by the three-letter codes of ISO 639-3, which aims to cover every known language. No new two-letter codes are issued.

Does Klingon really have an ISO code?

Yes, tlh. ISO 639-3 covers artificial languages alongside living and extinct ones. Esperanto, Volapük, Ido and Lojban are all registered too.

What is the difference between a language and a locale?

A language is a language. A locale is a language plus a region plus conventions for dates, currency and sorting. en is a language. en_GB is a locale. Unicode even defines a locale called root, for when a system has no idea.

What is a macrolanguage?

A code covering several closely related languages that are sometimes treated as one. Chinese, Arabic and Malay are macrolanguages. It exists because the boundary between a language and a dialect is a political question, not a linguistic one.

References

1. ISO 639, Language code
2. ISO 639 code tables, SIL International
3. ISO 639-3 released, SIL as registration authority
4. ISO 639-3