gobbledygook

This project is a procedural generator for fictional human languages.

It is a personal project developed as I discover interesting things about linguistics.

Key principles

Declarative approach

A Gobbledygook language is represented by a JSON object containing the rules, words, and specificities of that language

A system of rules and forms describes the way a word changes based on its context

A Gobbledygook sentence is represented by a JSON object that contains a semantic definition of that sentence as well as all relevant context information

Generating and Using a Gobbledygook language

Gobbledygook contains 2 main tools :

generate-language/ : creates a fictional languages with its rules and particularities, trying to make it sound interesting and organic
use-language/ : takes as input a language and a sentence definition to output the sentence in that language

A language that has been generated this way can be saved as a JSON and reused.

Language modelization principles

Gobbledygook assumes a very simple conception of language

Phonology

phoneme inventory, with frequencies
IPA representation and the language's transliteration code into latin alphabet (currently based on my personal preference rather than a recognized standard)

Morphology

Morphology types

Languages differ in how much words change based on context. Gobbledygook uses 3 simplified morphology types:

Type	Description	Example language	How "I love" changes
Flectional	Words change a lot, one suffix = multiple meanings	Latin	amo, amas, amat, amamus...
Semi-flectional	Some changes + some helper words	French	j'aime, tu aimes, il aime...
Analytical	Words barely change, meaning from word order + helpers	Mandarin	我爱 (wǒ ài) - no verb change

Morphemes

Morphemes are generated based on the above phoneme inventory by alternating vowels and consonants.

Declension system (nouns and adjectives)

A declension system describes how words change (or not) based on their context. The relevant parameters that inflect a word are listed as rules, and the word forms are described in a nested forms object.

The rules and forms system:

Rules list which parameters affect the word form, in order of nesting
Forms is a nested object following that order, ending with a template string

Example for English noun declension:

{
  rules: ["declensionType", "grammaticalCase", "number"],
  forms: {
    noun: {
      nominative: { singular: "{morpheme}", plural: "{morpheme}s" },
      genitive: { singular: "{morpheme}'s", plural: "{morpheme}s'" }
    },
    adjective: {
      nominative: { singular: "{morpheme}", plural: "{morpheme}" }
    }
  }
}

For "dog" (morpheme: "dog") in plural nominative: → navigate forms.noun.nominative.plural → "{morpheme}s" → "dogs"

More examples:

English inflects its nouns and adjectives by rules: ["declensionType", "grammaticalCase", "number"]

For yellow in the sentence the yellow taxis drive over the bridge, yellow follows the rules

declensionType = adjective,
grammaticalCase = nominative and
number = plural, giving the form yellow

French inflects its nouns and adjectives by rules: ["declensionType", "declensionGroup", "grammaticalCase", "gender", "number"]

For cheval in the sentence les enfants adorent les petits chevaux, cheval follows the rules

declensionType = noun,
declensionGroup = alEnding,
grammaticalCase = accusative,
gender = masculine and
number = plural, giving the form chevaux

Grammatical cases

Cases indicate the role a noun plays in a sentence. English mostly uses word order, but other languages change the word itself.

Case	Role	English example	Latin example
Nominative	Subject (who does the action)	"The dog bites the man"	"Canis mordet virum"
Accusative	Direct object (receives the action)	"The dog bites the man"	"Canis mordet virum"
Dative	Indirect object (recipient)	"I give a bone to the dog"	"Do os cani"
Genitive	Possession	"The dog's bone" / "bone of the dog"	"os canis"

In Latin, the word "canis" (dog) changes form: canis → canem → cani → canis depending on its role.

Semantic cases (locative cases)

Beyond grammatical roles, cases can express spatial and logical relationships. These are common in Finnish, Hungarian, and many other languages.

Case	Meaning	English equivalent	Example usage
Locative	Location (static)	"at", "in"	"I am at the house"
Lative	Movement towards	"to", "towards"	"I go to the house"
Inessive	Inside something	"inside", "within"	"The cat is in the box"
Ablative	Movement away from	"from", "out of"	"I come from the house"
Instrumental	Tool or means	"with", "using"	"I write with a pen"
Benefactive	Benefit recipient	"for"	"I made this for you"

In Finnish, "talo" (house) becomes: talossa (in the house), taloon (into the house), talosta (from the house).

Gobbledygook treats prepositions and case suffixes as equivalent mechanisms - both express the same semantic relationships.

Conjugation system (verbs)

The conjugation system follows the same rules and forms principles as declension.

Determiners and pronouns

Determination specifies which specific thing we're talking about:

Type	Meaning	English example	French example
Definite	A specific known thing	"the dog"	"le chien"
Indefinite	Any instance of a thing	"a dog"	"un chien"
Demonstrative	Pointing to something	"this/that dog"	"ce chien"
Possessive	Belonging to someone	"my/your/his dog"	"mon/ton/son chien"
Proper noun	A name (no determiner)	"Fido barks"	"Médor aboie"

The determiner and pronoun systems also follow the same rules and forms principles.

Syntax

A word order (SOV, SVO, VSO, VOS)
A phrase structure for: noun, verb, adjective
A sentence structure by type: declarative, interrogative (polar, open), negative

Lexicon

a dictionary of root morphemes with their grammatical properties (gender, declension or conjugation group, object category such as inanimate, animal, etc.)
abstract words for conjunctions adverbs and others (to be improved)

Simplifications and shortcuts

Gobbledygook cannot be exhaustive and represent all the possible languages. Some important simplications and shortcuts have been taken to make these tools.

Consonant clusters and diphtongs

They are represented as being a single vowel or consonant. They are created as such during language generation

Morphology types

I simplified the analytical - flectional continuum into 3 distinct categories to simplify the generation process

Grammatical and semantic cases are merged into one concept

Grammatical and semantic cases are represented by a single system that folds syntactic roles (nominative, accusative) with semantic relationships (inessive, benefactive)

Prepositions with semantic meaning are also meant to be folded in this system, as prefixes, stored in a dedicated object.

The system considers prepositions and suffix declensions are functionally the same, except with a space for prepositions.

example:

classic declension
- latin nominative {morpheme}us
- english genitive {morpheme}'s
preposition form as a declension form
- english allative {preposition} {morpheme} (/tu london/)
- french instrumental {preposition} {morpheme} (/avek panaʃ/)

This simplifies the representation and is sufficient to generate fictional languages.

Deterministic generation

One sentence definition will only output one output for one language.

Gobbledygook does not support style, synonymy or subtext

Absence is represented by an empty string

For example, absence of determiner in a language is represented with ""

A declension that hasn't a suffix will be represented simply "{morpheme}"

Real human languages are used as benchmark and development tests

I have used french, english and danish because I am familiar with these.

I also recognize that all 3 are closely related languages with close common ancestors and intense mutual influences, and are not super representative of the breadth of real life languages.

They are not fully implemented and many of their singularities are not represented.

These languages are meant to be used as tests for the useLanguage tool, and not exactly as an objective of the generateLanguage tool

Known limitations

All these limitations might become subject to future evolutions

Semantic expressivity accross languages

Idiomatic expressions and ways of expressing ideas varying between language can't be represented with Gobbledygook as it exists now.

If a language says a sentence like "The sky rains water drops" (transitive formulation) to say it rains, and another one says "it rains" (intransitive), Gobbledygook wouldn't be able to do that

If a language says "Il pleut comme vache qui pisse" (image), Gobbledygook can't represent this either.

Linguistic phenomena not represented

tone systems
agglutinative and polysynthetic languages
ergative-absolutive alignmed languages
tense aspects
negation
infixes, ablaut, reduplication
intonation and stress

Text to speech

Gobbledygook uses amazon's Polly to voice its languages. This model uses preset voices based on existing languages and accents. These voices are not able to produce all the sounds that Gobbledygook generates.

Future evolutions (big dreams)

Two new tools could be added to make this project extremely cool :

Language mutation generator

A tool that takes a language as input and a number of generations, and outputs a new language with realistic mutations.

Sentence encoder tool

A tool that takes in a human sentence (english probably) and encodes it into a Gobbledygook sentence definition.

LLMs might prove a good option, but there is probably a good traditional option.

Key project files

interfaces.ts : types for all manipulated concepts
generate-language/index.ts : generation tool entry point
use-language/make-sentence.ts : sentence production tool entry point
languages/french/index.ts : example of a real life language represented as a Gobbledygook language
sample-sentences/sample-sentences.ts : examples of sentence definitions in the Gobbbledygook format

Project vocabulary

Morphème : word root, base lexical unit
FormTable : structure { rules, forms } describing inflection of a word
rules : ordered list containing the inflection parameters (ex: ["gender", "number"])
forms : nested object, according to the rule orders, ending with leaf nodes with a template string for the word flection.
SentenceDefinition : Semantic representation of a sentence to transcribe
Context : parameter passed to the sentence builder, that contains the language and the consolidated sentence element and utterance context (speaker, addressee role, etc.)

Resources

Name		Name	Last commit message	Last commit date
Latest commit History 91 Commits
app		app
public		public
.eslintrc.json		.eslintrc.json
.gitignore		.gitignore
README.md		README.md
TODO.md		TODO.md
next-env.d.ts		next-env.d.ts
next.config.js		next.config.js
package.json		package.json
test.ts		test.ts
tsconfig.json		tsconfig.json
yarn.lock		yarn.lock

Folders and files

Latest commit

History

Repository files navigation

gobbledygook

Key principles

Declarative approach

Generating and Using a Gobbledygook language

Language modelization principles

Phonology

Morphology

Morphology types

Morphemes

Declension system (nouns and adjectives)

Grammatical cases

Semantic cases (locative cases)

Conjugation system (verbs)

Determiners and pronouns

Syntax

Lexicon

Simplifications and shortcuts

Consonant clusters and diphtongs

Morphology types

Grammatical and semantic cases are merged into one concept

Deterministic generation

Absence is represented by an empty string

Real human languages are used as benchmark and development tests

Known limitations

Semantic expressivity accross languages

Linguistic phenomena not represented

Text to speech

Future evolutions (big dreams)

Language mutation generator

Sentence encoder tool

Key project files

Project vocabulary

Resources

Language creation

IPA and phonetics

Morphology

Grammatical features

Language mutations

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Uh oh!

Contributors

Uh oh!

Languages

Packages