voynich-ipsum

Generate non-latinate random text from a conlang.

Summary

voynich-ipsum gives you a constructed language generator. You can specify various properties of the language -- it is synthetic or analytic; it is VSO or SVO or OSV; it has a simple or complex phoneme inventoty; etc. -- or leave this up to chance. The language is defined by the settings of various parameters and the construction of various probability distributions during instantiation. After this it has a fixed name and properties. You can ask it to generate random text with our without a list of topics.

Mostly this is intended as a replacement for boring, repetitive lorem ipsum text.

Why "voynich-ipsum"?

The Voynich Manuscript is a famous example of a text that is maybe real, maybe a fraud. Examining the statistical properties of the characters, words, and structure of the text suggests mostly that is has the structure of a real language, but not entirely, and any affiliation with an existing language and/or writing system is still up in the air.

The aim with voynich-ipsum is to generate explicitly fake languages with the same properties. The code incorporates many real linguistic properties to produce a patina of plausibility for text that means nothing at all.

Synopsis

import { Language } from 'voynich-ipsum'

const lang = new Language({seed: 1})
console.log(lang.name)
// => Ekogo

// produce a 5-paragraph essay on a random topic
console.log(lang.essay(5))
// => 
// Ezozsebe azeghi ze. Ghoxeexl sedgog ghudlsooz. Ghodoghi aso akuoed do buoghupe.
//
// Saxeexl zaib adiupu.
//
// Uea ghuzazseu.
//
// Kakoko uvbnsu. Kakoko ghudili. Akoko zo e ghuipu be.
//
// Kakoko ghusodobabo. Sazope aakupo.

// assert something about 'foo'
console.log(lang.assertion('foo'))
// => Ufoo adiode.

// ask a question about 'foo'
console.log(lang.question('foo'))
// => Efoo ovde?

// shout something about 'foo'
console.log(lang.exclamation('foo'))
// => Foo doeeavuzl zeze!

// a phrase concerning foo
console.log(lang.nounPhrase('foo'))
// => afoo

Please look at the source code for more guidance. The public methods are documented.

Caveats

The choices in here are based in some cases on published research in linguistic typology, sometimes in statistics I scraped out of random text I found on the Internet, and most often out of plausible-seeming numbers I pulled out of my ass. There is a vast amount of linguistic variation I have not covered. This does not produce ergative-absolutive languages (or nominative-accusative languages, for that matter). It does not know about vowel harmony. It can't produce languages with tone, or nasal vowels. It can't do noun class agreement between nouns and adjectives. It doesn't really know about adjectives at all. It produces phonemes that were easy to produce with a mostly latinate character set. It has no variation in orthography. It never produces a right-to-left language. Etc. Etc. Etc.

I had great ambitions for covering typological variables and gave up on most of them because I'm lazy and my attention wanders.

Sub-caveat

Because parameters are set by a random number generator during initialization, code changes are liable to completely change the language generated with a particular random number seed. If this matters to you, you should specify a particular version number in your dependencies so your languages won't change dramatically if a new version of lorem-voynich is released. If you're just happy to get random gibberish, don't worry about this.

Contributing

If you are less lazy or more knowledgeable than me, send me pull requests! Because I am lazy and my attention wanders, I may not respond to them promptly. I apologize for that in advance.

Acknowledgements

I would like to thank my wife and kids, who put up with too much of my nonsense and random enthusiasms.

I would also like to thank my employer, Green River, who also show much patience for my vagaries.

This was all done on my own time for reasons which are not clear even to me.